1025-00008 · Quorum Insurance Holdings · Transaction Documents/MDLA

derived-data-rights-revised-compromise.docx DOCUMENT M&A / Corporate

61 / 76 ← → ✕

PRIVILEGED & CONFIDENTIAL — ATTORNEY WORK PRODUCT — DRAFT FOR DISCUSSION PURPOSES ONLY

REVISED ARTICLE VII — DERIVED DATA AND INTELLECTUAL PROPERTY RIGHTS: QUORUM COMPROMISE FRAMEWORK

Proposed Revisions to Master Data License Agreement Between Quorum Insurance Holdings Corp. and Meridian Credit intelligence, Inc.

Date: June 2023

Matter No.: C&H Matter No. 1025-00008

Prepared by: Wendy Schultz, Associate, intellectual Property, Calderwood & Harkness LLP

With contributions from:

• Omar Siddiqui, Associate, intellectual Property, Calderwood & Harkness LLP

• Patricia Heng, Senior Counsel, Technology Transactions, Calderwood & Harkness LLP

NOTICE: This document is protected by the attorney-client privilege and the work product doctrine. It is intended solely for the use of the client, Quorum Insurance Holdings Corp., and is being circulated within Calderwood & Harkness LLP and to designated Quorum personnel for purposes of legal advice in connection with the negotiation of the Master Data License Agreement with Meridian Credit intelligence, Inc. This document should not be disclosed, distributed, or disseminated to any third party without the prior written consent of Calderwood & Harkness LLP.

Note: This document sets forth Quorum Insurance Holdings Corp.'s ("Quorum") revised proposed Article VII provisions for the Master Data License Agreement ("MDLA") with Meridian Credit intelligence, Inc. ("Meridian"), reflecting a compromise framework developed in advance of the June 30, 2023 third negotiation session in New York. The proposed provisions herein are intended to bridge the impasse over derived-data rights that has persisted since the first in-person negotiation session and to present a commercially reasonable framework that protects Quorum's core interests while acknowledging Meridian's legitimate concerns regarding its raw data assets.

Right-click to update Table of Contents

I. INTRODUCTORY MEMORANDUM / COVER NOTE

A. Purpose and Context

This revised Article VII represents Quorum's good-faith attempt to bridge the impasse over derived-data rights that has persisted since the January 18, 2023 first in-person negotiation session and remained unresolved through the March 9, 2023 second session in Atlanta. The parties have exchanged multiple marked-up drafts, engaged in extensive telephonic discussions between counsel, and have made substantial progress on many other articles of the MDLA. Article VII, however, remains the single most commercially significant open issue in the agreement, and its resolution is a prerequisite to closing the transaction on the timeline contemplated by both parties' respective executive teams.

Quorum's Original Position (January 2023 Draft of Article VII). In its initial draft circulated on January 12, 2023, Quorum proposed that Quorum own all intellectual property in Derived Data with unrestricted rights to use, sublicense, and commercialize such Derived Data without additional fees, royalty obligations, or Meridian consent. Under Quorum's original formulation, Derived Data was treated as Quorum's proprietary work product from the moment of creation, and Meridian had no continuing interest in, or rights with respect to, any analytical outputs generated by Quorum's processing of the Licensed Data. This position reflected Quorum's view — which we continue to believe is correct as a matter of law — that the analytical outputs produced by Quorum's proprietary algorithms, combined with Quorum's First-Party Data, represent independent works of authorship and invention that do not infringe or implicate Meridian's intellectual property rights.

Meridian's Counter-Position (November 2022 Form MDLA and Subsequent Exchanges). Meridian, through its outside counsel at Kessler Pruitt & Ashe LLP (Graham Whitley, Nina Chastain, and Derek Fontaine), advanced a dramatically different framework in the November 4, 2022 form MDLA and in subsequent exchanges. Meridian's position can be summarized as follows:

(a) Derived Data constitutes a "derivative work" under the Copyright Act (17 U.S.C. § 101 et seq.) and is therefore subject to Meridian's underlying intellectual property rights in the Licensed Data;

(b) Meridian demands a perpetual, royalty-free license-back on all Derived Data, granting Meridian the right to use, copy, modify, and distribute Quorum's Derived Data for Meridian's own commercial purposes;

(c) Quorum would be prohibited from any third-party commercialization of Derived Data unless Meridian determines, in its sole discretion, that the Derived Data reflects "sufficient transformation" of the Licensed Data to warrant independent treatment; and

(d) On those occasions where Meridian approves third-party commercialization, Meridian demands a 15% revenue share on net revenues attributable to the commercialized Derived Data.

Quorum views each of these demands as commercially unreasonable and legally unsupported. The license-back demand would effectively allow Meridian to free-ride on Quorum's multi-million-dollar investment in data science, machine learning infrastructure, and proprietary algorithm development. The sole-discretion transformation standard would give Meridian an effective veto over Quorum's commercial strategy. And the 15% revenue share is untethered to any industry benchmark or reasonable allocation of value, particularly given that Quorum's Derived Data reflects the combined contribution of Quorum's First-Party Data, Quorum's proprietary technology, and the Licensed Data.

This Compromise Framework. This revised Article VII moves Quorum from its maximalist position while firmly rejecting Meridian's core demands for a license-back, sole-discretion transformation standard, and revenue share. The key principles of the compromise are as follows:

(1) Quorum Retains Full Ownership of Derived Data. As between the parties, Quorum shall own all right, title, and interest in and to all Derived Data, consistent with the well-established principle that licensees own the work product they create using licensed inputs.

(2) No License-Back to Meridian. Quorum does not grant, and shall not be deemed to have granted, any license or right of access to Meridian in or to Quorum's Derived Data, Model Inference Outputs, or proprietary models and algorithms.

(3) No Revenue Share to Meridian. Quorum shall not owe Meridian any royalty, revenue share, profit participation, or other compensation in connection with Quorum's commercialization of Derived Data.

(4) Objective Transformation Standard for Third-Party Commercialization. Quorum may commercialize Derived Data to third parties provided the outputs meet an objective, jointly agreed de-identification and transformation standard — replacing Meridian's demanded sole-discretion standard with a framework based on measurable criteria and an independent Technical Neutral dispute resolution mechanism.

(5) Specific Protection for QuorumScore™. Outputs of Quorum's proprietary QuorumScore™ predictive underwriting model are pre-certified as Sufficiently Transformed Outputs, exempting them from case-by-case review and protecting Quorum's planned B2B licensing revenue line.

We believe this framework represents a commercially reasonable middle ground that protects Meridian's legitimate interest in preventing redistribution of its raw data while recognizing Quorum's substantial independent investment in analytics and proprietary technology. The framework is consistent with market practice in comparable data-licensing transactions and provides Meridian with meaningful protections — including an objective transformation standard, a notification and challenge mechanism, and limited audit rights — that Quorum's original draft did not include.

B. Relationship to Other Negotiation Documents

This revised Article VII should be read in conjunction with the following related documents and workstreams:

Quorum's Original Proposed Article VII (January 2023). The January 12, 2023 draft of Article VII, as circulated to Kessler Pruitt & Ashe on that date, reflects Quorum's initial position. The present document supersedes that draft in its entirety and should be treated as the operative Quorum proposal for the upcoming June 30, 2023 negotiation session.

Draft Derived Data Commercialization Protocol (Appendix VII-A). A companion document, the Derived Data Commercialization Protocol, is being developed concurrently by Wendy Schultz and Omar Siddiqui as a practical operational supplement to the proposed Article VII. The Protocol will set forth the specific format and content requirements for Transformation Certificates, detailed procedures for the Technical Neutral selection and review process, and illustrative examples of outputs that would and would not qualify as Sufficiently Transformed Outputs. The Protocol is intended to be appended to the MDLA as Appendix VII-A but is designed to be amendable by mutual written agreement of the parties without a formal MDLA amendment, providing operational flexibility as the parties' practices evolve.

Pinnacle Ridge Advisory LLC Pricing Benchmark Study. The pricing benchmark study delivered by Pinnacle Ridge Advisory LLC (Franklin Osei, Managing Director) on February 14, 2023 informs the commercial rationale underlying this compromise framework. The Pinnacle Ridge study surveyed derived-data rights provisions in 23 comparable data-licensing transactions across the financial services and insurance sectors completed between 2019 and 2022. The study's principal finding — that licensee ownership of derived data is the prevailing market norm, with revenue-share provisions appearing in fewer than 15% of surveyed transactions — supports Quorum's position that ownership without revenue sharing is the commercially reasonable baseline.

Most Favored Nation Pricing Issue. The most favored nation ("MFN") pricing issue remains separately unresolved and is not addressed in this document. This document addresses only derived-data rights under Article VII. We note, however, that progress on the derived-data issue may create goodwill and momentum for resolution of the MFN pricing dispute in subsequent sessions.

C. Instruction to Counsel

This draft is formatted as proposed contract language for Article VII of the MDLA, with drafting notes and commentary included in bracketed italicized text for internal Calderwood & Harkness and Quorum review purposes. All such drafting notes and internal commentary will be removed prior to transmission of the proposed Article VII to Kessler Pruitt & Ashe LLP.

This document is intended for discussion with Carolyn Whitford, General Counsel of Quorum Insurance Holdings Corp., and Gregory Stahl, Vice President, Legal Affairs, of Quorum Insurance Holdings Corp., prior to any transmission to opposing counsel. Upon approval by Ms. Whitford and Mr. Stahl, the proposed Article VII (stripped of internal drafting notes) will be transmitted to Kessler Pruitt & Ashe LLP (Graham Whitley, Nina Chastain, and Derek Fontaine) in advance of the June 30, 2023 negotiation session, with a cover letter from Patricia Heng framing the proposal as a compromise framework and requesting that Kessler Pruitt & Ashe review and respond with a marked-up version or counter-proposal no later than June 23, 2023.

All internal Calderwood & Harkness team members should direct questions regarding this draft to Wendy Schultz (ext. 4218) or Patricia Heng (ext. 3802).

II. PROPOSED ARTICLE VII: DERIVED DATA AND INTELLECTUAL PROPERTY RIGHTS

[Note: The following constitutes the proposed contract language for Article VII of the MDLA. Bracketed italicized text represents internal drafting notes and commentary for Calderwood & Harkness and Quorum review only and will be removed from the version transmitted to Kessler Pruitt & Ashe LLP.]

ARTICLE VII

DERIVED DATA AND INTELLECTUAL PROPERTY RIGHTS

Section 7.1 — Definitions

For purposes of this Article VII, the following capitalized terms shall have the meanings set forth below. Capitalized terms used but not defined in this Article VII shall have the meanings ascribed to them elsewhere in this Agreement.

(a) "Derived Data" means any and all data, datasets, databases, indices, scores, risk assessments, models, algorithms, model weights, model parameters, statistical outputs, analytics results, reports, blended risk scores, propensity models, enriched datasets, and other works or materials created, generated, compiled, developed, or produced by or on behalf of Licensee or its Affiliates using, incorporating, referencing, or processing the Licensed Data (as defined in Article II) as an input, whether in whole or in part, and whether alone or in combination with Licensee's own First-Party Data or third-party data. For the avoidance of doubt, "Derived Data" includes, without limitation:

(i) blended risk scores that combine Licensed Data (including Meridian credit data elements) with Licensee's First-Party Data, including but not limited to Licensee's proprietary claims data, policy data, and telematics data;

(ii) propensity models developed using Licensed Data as a training input, calibration input, or validation input;

(iii) enriched datasets created by appending, merging, or integrating Licensed Data with Licensee's First-Party Data or with data from other third-party sources;

(iv) outputs of machine-learning models, artificial intelligence systems, neural networks, gradient-boosted decision trees, ensemble models, and other statistical or computational models that were trained, calibrated, validated, or otherwise developed using Licensed Data as an input, whether in whole or in part; and

(v) outputs of QuorumScore™ (Licensee's proprietary predictive underwriting model, United States Patent and Trademark Office application pending, Serial No. 97/482,116), including all scores, risk assessments, classification outputs, and predictive analytics outputs generated by QuorumScore™.

(b) "First-Party Data" means data owned, collected, generated, or licensed by Licensee independently of the Licensed Data, including, without limitation, Licensee's proprietary claims data, policyholder data, telematics data, loss-run data, agent and broker data, actuarial data, premium data, and any data acquired by Licensee from sources other than Licensor. For the avoidance of doubt, First-Party Data does not include any data derived from or incorporating the Licensed Data.

(c) "Licensed Data" shall have the meaning ascribed to such term in Section 2.1 of this Agreement, which encompasses, without limitation, the following Licensor data products and datasets:

(i) MCI Consumer Credit File™, consisting of approximately 310 million individual consumer credit records maintained by Licensor;

(ii) MCI CommercialView™, consisting of approximately 42 million commercial business credit records maintained by Licensor; and

(iii) MCI AltSignal™, consisting of approximately 85 million individual consumer alternative-data records maintained by Licensor, including utility payment data, rental payment data, telecommunications payment data, and other non-traditional credit indicators.

The specific scope, format, delivery schedule, and update frequency for the Licensed Data are set forth in Article II and the Data Delivery Specifications attached as Appendix II-A.

(d) "Sufficiently Transformed Output" means Derived Data that has been de-identified and transformed such that:

(i) no individual record of Licensed Data can be extracted, reconstructed, reverse-engineered, or re-identified from the Derived Data through commercially reasonable means;

(ii) the Derived Data does not contain, display, or transmit any raw Licensed Data fields, tradeline-level data, or consumer-level or business-level records attributable to the Licensed Data; and

(iii) the Derived Data represents a materially new and independent work reflecting Licensee's proprietary analytical methodologies, algorithms, and/or First-Party Data contributions.

The determination of whether particular Derived Data constitutes a "Sufficiently Transformed Output" shall be made solely by reference to the objective criteria set forth in Section 7.5(a) and the procedures set forth in the Derived Data Commercialization Protocol (Appendix VII-A), and shall not be subject to either party's sole or unilateral discretion.

[Drafting Note: This definition, and in particular the proviso regarding objective criteria rather than sole discretion, is the central innovation of the compromise framework. Meridian's November 2022 form MDLA proposed that Meridian would determine "sufficient transformation" in its sole discretion — a standard that would effectively give Meridian veto power over Quorum's commercial strategy. By replacing Meridian's sole-discretion standard with objective, measurable criteria and an independent Technical Neutral dispute resolution process, we preserve Quorum's autonomy while providing Meridian with a meaningful and transparent framework for protecting its raw data.]

(e) "Model Inference Outputs" means predictions, scores, classifications, probability estimates, risk assessments, or other outputs generated by a trained machine-learning model, statistical model, or other computational model during live inference (i.e., the application of the trained model to new input data for purposes of generating predictions or other outputs), where the model was trained, calibrated, or validated using Licensed Data (among other inputs) but the live inference inputs do not include Licensed Data. For the avoidance of doubt, Model Inference Outputs include outputs generated by models that have "learned" statistical patterns, relationships, or correlations from Licensed Data during the training process but that do not retain, store, or transmit the underlying Licensed Data records used during training.

[Drafting Note: This definition is critical to resolving the technical dispute that arose during the March 9 session regarding whether ML model outputs constitute "derived data." Kessler Pruitt & Ashe's position — advanced by Derek Fontaine in his March 15 letter — is that any output from a model trained on Meridian data is inherently "derived" from that data and therefore subject to Meridian's derived-data restrictions. Quorum's position, which is technically and legally correct, is that Model Inference Outputs generated from non-Meridian inputs are independent outputs. During live inference, the model applies learned statistical relationships (encoded in model weights and parameters) to new data; the underlying training data is not accessed, referenced, or transmitted. This is analogous to a human analyst who has studied credit data and then applies learned expertise to new, independent data — the analyst's conclusions are not "derived from" the training data in any legally cognizable sense. We should be prepared to support this position with expert testimony from Dr. Ananya Krishnamurthy (MIT CSAIL) if Meridian escalates this issue.]

Section 7.2 — Ownership of Derived Data

(a) Licensee Ownership. As between the parties, Licensee shall own all right, title, and interest (including all intellectual property rights, whether arising under copyright, patent, trade secret, database rights, or other applicable law) in and to all Derived Data created, generated, compiled, developed, or produced by or on behalf of Licensee or its Affiliates during the Term or at any time thereafter.

Licensor acknowledges and agrees that Derived Data does not constitute a "derivative work" (as that term is defined in 17 U.S.C. § 101 et seq. and as interpreted by applicable case law) of the Licensed Data, and that Licensor retains no intellectual property rights in or to Derived Data by virtue of Licensor's ownership of or intellectual property rights in the Licensed Data. The parties acknowledge that the creation of Derived Data involves substantial independent intellectual contribution by Licensee — including the application of Licensee's proprietary algorithms, methodologies, trade secrets, and First-Party Data — and that the resulting Derived Data constitutes an original and independent work of Licensee.

To the extent that any Derived Data is deemed under applicable law, by any court of competent jurisdiction or other governmental authority, to constitute a derivative work of the Licensed Data or otherwise to be subject to Licensor's intellectual property rights, Licensor hereby irrevocably assigns, transfers, and conveys to Licensee all of Licensor's right, title, and interest in and to such Derived Data, and agrees to execute, at Licensee's request and expense, such further documents and instruments as may be reasonably necessary to evidence, perfect, and record such assignment.

(b) Reservation of Rights in Licensed Data. Nothing in this Article VII shall be construed to transfer to Licensee any ownership interest in the Licensed Data itself. Licensor retains all right, title, and interest in and to the Licensed Data, including all intellectual property rights therein, as set forth in Article II of this Agreement. Licensee's rights to access, use, copy, process, and otherwise exploit the Licensed Data are governed solely by the license grant set forth in Article III of this Agreement and are subject to all restrictions, limitations, and conditions set forth therein and elsewhere in this Agreement.

(c) No License-Back. Licensee does not grant, and shall not be deemed to have granted, to Licensor or any Licensor Affiliate any license, sublicense, right of access, right of use, or other right or interest in or to:

(i) any Derived Data;

(ii) any Model Inference Outputs;

(iii) any of Licensee's proprietary models, algorithms, model architectures, model weights, model parameters, source code, object code, training methodologies, or data-processing pipelines;

(iv) QuorumScore™ or any outputs thereof; or

(v) any other intellectual property, trade secrets, confidential information, or proprietary materials of Licensee or its Affiliates.

Nothing in this Agreement shall be interpreted as an implied license, license by estoppel, or other grant of rights from Licensee to Licensor in any of the foregoing.

[Drafting Note: This provision directly and categorically rejects Meridian's demand, set forth in Section 7.3(b) of Meridian's November 2022 form MDLA, for a "perpetual, irrevocable, worldwide, royalty-free, fully paid-up, sublicensable license" to all Derived Data. The license-back demand is commercially unacceptable because it would effectively grant Meridian the benefit of Quorum's multi-million-dollar investment in data science without any corresponding compensation or investment by Meridian. We anticipate strong pushback from Kessler Pruitt & Ashe on this point and should prepare Carolyn Whitford for the discussion.]

(d) Model Inference Outputs. All Model Inference Outputs shall be owned exclusively by Licensee, free and clear of any claims, rights, encumbrances, or interests of Licensor. Model Inference Outputs shall not constitute Derived Data for purposes of this Agreement and shall not be subject to any restrictions, limitations, obligations, or conditions under this Article VII or elsewhere in this Agreement. Without limiting the generality of the foregoing:

(i) Model Inference Outputs generated using non-Licensor inputs shall not be subject to the commercialization provisions of Section 7.4 or the transformation standard and certification process of Section 7.5;

(ii) Licensee may use, license, sublicense, sell, distribute, and otherwise exploit Model Inference Outputs for any purpose, including third-party commercialization, without notice to, consent of, or payment to Licensor; and

(iii) Licensee's right to use and exploit Model Inference Outputs shall survive expiration or termination of this Agreement in perpetuity.

[Drafting Note: This provision resolves the technical dispute over ML model outputs squarely in Quorum's favor. During live inference, a trained model applies learned statistical patterns to new input data; the training data (including Licensed Data) is not accessed, referenced, or retransmitted. The model weights and parameters are mathematical representations of learned correlations — they do not "contain" the underlying training data in any meaningful sense. This is consistent with the emerging consensus in the data-licensing industry and is supported by the technical analysis prepared by Quorum's Chief Data Officer, Dr. Sanjay Mehrotra. If Meridian escalates this issue, we should consider engaging an independent expert to provide a technical opinion supporting this position.]

Section 7.3 — Licensee's Internal Use Rights

(a) Unrestricted Internal Use. Licensee shall have the unrestricted, perpetual (subject to the Term limitations on creating new Derived Data set forth in Section 7.7(b)), irrevocable, fully paid-up, worldwide right to use, copy, store, modify, adapt, enhance, analyze, reprocess, reformat, and create further derivatives of Derived Data for any and all internal business purposes. Without limiting the generality of the foregoing, such internal business purposes include:

(i) automated underwriting and risk assessment across Licensee's property and casualty, life, specialty, and reinsurance lines of business;

(ii) claims-fraud detection, prevention, and investigation, including the development and deployment of machine-learning models for real-time fraud scoring;

(iii) customer propensity scoring, lifetime value modeling, cross-sell and upsell targeting, and customer segmentation for marketing and distribution purposes;

(iv) actuarial analysis, including loss reserving, rate-making, experience studies, and catastrophe modeling;

(v) regulatory reporting, compliance monitoring, and responses to inquiries or examinations by state insurance departments, the National Association of Insurance Commissioners, or other regulatory authorities;

(vi) internal research and development, including the development, training, testing, validation, and refinement of new and existing predictive models, algorithms, and analytical tools;

(vii) enterprise risk management, including portfolio analysis, concentration risk assessment, and reinsurance purchasing decisions; and

(viii) any other lawful internal business purpose as determined by Licensee in its sole discretion.

(b) Affiliate Use. Licensee may make Derived Data (including Sufficiently Transformed Outputs and Model Inference Outputs) available to its Affiliates for internal business purposes on the same terms and subject to the same conditions as apply to Licensee under this Article VII, without additional fee, royalty, or compensation to Licensor and without any requirement for prior notice to or consent from Licensor. For purposes of this Section 7.3(b), "Affiliates" shall have the meaning ascribed to such term in Section 1.1 of this Agreement, and each Affiliate receiving Derived Data shall be bound by the confidentiality obligations and data security requirements applicable to Licensee under this Agreement.

(c) QuorumScore™ Internal Use. Without limiting the generality of Sections 7.3(a) and 7.3(b), Licensee's use of QuorumScore™ (USPTO application Serial No. 97/482,116) and all outputs thereof for Licensee's internal underwriting, pricing, risk-management, and other internal business purposes shall be unrestricted and shall not require any notice to, consent of, or payment to Licensor, regardless of the extent to which Licensed Data was used as a training input, calibration input, or validation input in the development or refinement of QuorumScore™.

Section 7.4 — Commercialization of Derived Data to Third Parties

(a) Permitted Third-Party Commercialization. Subject to the conditions set forth in this Section 7.4 and Section 7.5, Licensee shall have the right to license, sublicense, sell, distribute, publish, or otherwise make available Derived Data to third parties (including entities outside Licensee's corporate group) for commercial purposes, provided that such Derived Data constitutes a "Sufficiently Transformed Output" as defined in Section 7.1(d) and as determined pursuant to the objective criteria and procedures set forth in Section 7.5 and the Derived Data Commercialization Protocol attached hereto as Appendix VII-A.

For the avoidance of doubt, Licensee's right to commercialize Derived Data under this Section 7.4 extends to any lawful commercial use, including, without limitation, licensing of scores, risk assessments, and analytical outputs to insurers, reinsurers, financial institutions, healthcare organizations, government agencies, and other entities, whether on a subscription basis, per-query basis, or any other pricing model determined by Licensee in its sole discretion.

(b) No Revenue Share. Licensee shall not owe Licensor any royalty, revenue share, profit participation, milestone payment, or other compensation of any kind in connection with Licensee's commercialization, licensing, sale, distribution, or other exploitation of Derived Data, Sufficiently Transformed Outputs, or Model Inference Outputs to third parties. All revenues, profits, and other economic benefits arising from such commercialization shall belong exclusively to Licensee.

[Drafting Note: This provision directly and categorically rejects Meridian's demand, set forth in Section 7.5(d) of Meridian's November 2022 form MDLA, for a 15% revenue share on net revenues attributable to third-party commercialization of Derived Data. The revenue share demand is commercially unreasonable for several reasons: (i) it is not supported by market practice, as confirmed by the Pinnacle Ridge Advisory study; (ii) it fails to account for Quorum's substantial independent investment in technology, data science talent, and First-Party Data; (iii) it would create perverse incentives for Meridian to inflate the scope of "Derived Data" to maximize its revenue share; and (iv) it is inconsistent with the parties' agreed license fee structure under Article VI, which already compensates Meridian for Quorum's use of the Licensed Data. We should be prepared to articulate these points clearly at the June 30 session.]

(c) QuorumScore™ Commercialization Carve-Out. Outputs of QuorumScore™ (including, without limitation, all scores, risk assessments, classifications, probability estimates, and predictive analytics outputs generated by QuorumScore™) shall be deemed Sufficiently Transformed Outputs for all purposes of this Section 7.4 and this Article VII, without further review, certification, or determination under Section 7.5, provided that such outputs do not contain or transmit any raw Licensed Data fields, tradeline-level data, or consumer-level or business-level records directly attributable to the Licensed Data.

For the avoidance of doubt, QuorumScore™ outputs that are numerical scores, probabilistic assessments, risk classifications, or similar analytical outputs — as distinguished from raw or lightly processed Licensed Data — shall be presumed to satisfy the foregoing condition, and Licensor shall bear the burden of demonstrating that any particular QuorumScore™ output fails to satisfy such condition.

[Drafting Note: The QuorumScore™ carve-out is essential to protect Quorum's competitive position and its planned B2B licensing revenue line. QuorumScore™ (USPTO application Serial No. 97/482,116) represents Quorum's single most significant proprietary technology asset, blending Meridian credit data with Quorum's own first-party claims, policy, and telematics data through proprietary machine-learning algorithms (including gradient-boosted decision trees and deep neural network ensembles). The resulting scores represent Quorum's independent intellectual contribution and are central to Quorum's planned B2B licensing business, which is projected to generate $40–50 million in annual revenue by 2027 based on internal forecasts reviewed by Quorum's Board of Directors. Subjecting QuorumScore™ outputs to a case-by-case Transformation Certificate process would create unacceptable commercial uncertainty and could delay or impede Quorum's go-to-market strategy. This carve-out should be treated as a non-negotiable element of the compromise framework; Quorum should be prepared to walk away from the MDLA if Meridian insists on subjecting QuorumScore™ to case-by-case review or revenue sharing.]

(d) Restrictions on Raw Data Distribution. For the avoidance of doubt, nothing in this Article VII permits Licensee to distribute, sublicense, sell, or otherwise make available to any third party:

(i) the Licensed Data in raw or unprocessed form;

(ii) any dataset that contains raw Licensed Data fields, tradeline-level data, or consumer-level or business-level records directly attributable to the Licensed Data, including, without limitation, individual credit scores generated by Licensor, individual payment histories, account balances, credit limits, delinquency statuses, or other data elements sourced from MCI Consumer Credit File™, MCI CommercialView™, or MCI AltSignal™ in an identifiable or record-level form; or

(iii) any output, report, dataset, or other material from which raw Licensed Data records can be extracted, reconstructed, or reverse-engineered through commercially reasonable technical means, including, without limitation, through the application of record linkage techniques, statistical disclosure attacks, or other re-identification methodologies that are commercially available or widely known in the data science community as of the date of distribution.

Any distribution of Licensed Data to third parties is governed exclusively by the terms and restrictions of Article III and, if applicable, Article IV (Permitted Disclosures), and nothing in this Article VII expands or modifies Licensee's rights under those Articles.

Section 7.5 — Objective Transformation Standard and Procedures

(a) Objective Criteria. Whether Derived Data constitutes a "Sufficiently Transformed Output" for purposes of Section 7.4 shall be determined solely by reference to the following objective criteria, and not by either party's sole, unilateral, or unfettered discretion:

(i) No Extractability. Individual consumer records or business records from the Licensed Data cannot be extracted, reconstructed, or re-identified from the Derived Data through commercially reasonable technical means. For purposes of this criterion, "commercially reasonable technical means" means technical methods, tools, and approaches that are commercially available, widely known in the data science and privacy engineering communities, and economically feasible to deploy, and does not include theoretical attacks requiring extraordinary computational resources, access to non-public information, or expertise materially beyond the state of the art.

(ii) No Raw Data Inclusion. The Derived Data does not contain, display, transmit, or make accessible any raw Licensed Data fields, including, without limitation, individual tradeline data, payment histories, account balances, credit limits, delinquency statuses, inquiry records, or identifying information (such as names, Social Security numbers, dates of birth, or addresses) sourced from MCI Consumer Credit File™, MCI CommercialView™, or MCI AltSignal™.

(iii) Material Transformation. The Derived Data reflects material analytical, statistical, algorithmic, or computational transformation by Licensee, incorporating Licensee's proprietary methodologies, algorithms, trained model parameters, and/or First-Party Data, such that the Derived Data constitutes a materially new and independent work that is qualitatively and substantively different from the Licensed Data used as an input. The mere reformatting, reorganizing, sorting, filtering, or subsetting of Licensed Data, without material analytical or computational transformation, shall not constitute a Sufficiently Transformed Output.

(iv) Aggregation/Anonymization Threshold. To the extent the Derived Data includes aggregate or statistical outputs (such as summary statistics, cohort-level risk assessments, geographic risk indices, or industry trend analyses), such outputs shall be derived from a minimum cohort of fifty (50) individual records and shall not permit the identification, whether directly or through linkage with other reasonably available information, of any individual consumer or business entity.

[Drafting Note: We have proposed a minimum cohort of 50 records for the aggregation/anonymization threshold. This number is consistent with the HIPAA Safe Harbor de-identification standard (which uses a minimum cell size approach) and with common practice in the statistical disclosure control literature. Meridian may push for a higher threshold (e.g., 100 or 250 records). We believe 50 is defensible and commercially reasonable. The specific number is bracketed as a potential negotiation point, but we should not go below 50.]

(b) Certification Process. Prior to the first commercial distribution to a third party of any new category of Derived Data (other than QuorumScore™ outputs, which are pre-certified as Sufficiently Transformed Outputs under Section 7.4(c) and are not subject to this Section 7.5(b)), Licensee shall deliver to Licensor a written certification (a "Transformation Certificate") that includes the following:

(i) a general description of the category of Derived Data to be commercialized, including the nature of the analytical outputs, the intended use cases, and the types of third-party recipients;

(ii) a written confirmation, signed by a duly authorized officer of Licensee, that the Derived Data meets each of the objective criteria set forth in Section 7.5(a)(i) through (iv); and

(iii) a description of the transformation methodology applied to the Licensed Data at a level of detail sufficient to demonstrate compliance with the criteria in Section 7.5(a), provided that Licensee shall not be required to disclose its proprietary algorithms, model architectures, source code, trade secrets, or other confidential information beyond the level of detail reasonably necessary to demonstrate compliance.

For the avoidance of doubt, the Transformation Certificate is a notification mechanism. Licensee's right to commercialize Sufficiently Transformed Outputs is not contingent upon Licensor's approval, consent, acknowledgment, or non-objection. The delivery of a Transformation Certificate is a condition to Licensee's right to commence commercial distribution of the applicable category of Derived Data, but Licensor's failure to respond to or acknowledge receipt of a Transformation Certificate shall not impair Licensee's rights hereunder.

(c) Licensor's Right to Challenge. Licensor may, within thirty (30) calendar days following Licensor's receipt of a Transformation Certificate (the "Challenge Period"), deliver to Licensee a written objection (a "Transformation Challenge") setting forth in reasonable detail:

(i) the specific criterion or criteria in Section 7.5(a) that Licensor believes the Derived Data does not satisfy; and

(ii) the factual and technical basis for Licensor's belief, including, where applicable, a description of the technical methodology by which Licensor contends that Licensed Data could be extracted, reconstructed, or re-identified from the Derived Data.

A Transformation Challenge that fails to identify a specific criterion in Section 7.5(a) and provide a reasonable factual and technical basis for the challenge shall be deemed insufficient and shall not constitute a valid objection under this Section 7.5(c).

If Licensor does not deliver a valid Transformation Challenge within the Challenge Period, the Derived Data described in the applicable Transformation Certificate shall be irrevocably deemed a Sufficiently Transformed Output for all purposes of this Agreement, and Licensee shall be free to commence and continue commercial distribution of such Derived Data without further notice to or involvement by Licensor.

(d) Dispute Resolution for Transformation Disputes. If Licensor delivers a valid and timely Transformation Challenge under Section 7.5(c), the following dispute resolution procedures shall apply:

(i) Good-Faith Negotiation. The parties shall first attempt to resolve the dispute through good-faith negotiation between designated representatives of each party (at the senior vice president level or above) for a period of thirty (30) calendar days following Licensee's receipt of the Transformation Challenge (the "Negotiation Period"). Each party shall make its designated representative reasonably available for meetings and calls during the Negotiation Period, and the parties shall exchange relevant technical information on a confidential basis to facilitate resolution.

(ii) Technical Neutral. If the dispute is not resolved through good-faith negotiation within the Negotiation Period, either party may, by written notice to the other party, submit the dispute to an independent technical expert (a "Technical Neutral") for binding determination. The Technical Neutral shall be a qualified data-science professional with at least fifteen (15) years of experience in data analytics, data privacy, statistical disclosure control, or a closely related field, and shall not have any material conflict of interest with either party. The parties shall use good-faith efforts to agree upon the selection of the Technical Neutral within fifteen (15) calendar days of the notice submitting the dispute. If the parties are unable to agree upon the selection of the Technical Neutral within such fifteen (15) day period, either party may request that the American Arbitration Association appoint the Technical Neutral from its panel of qualified data-science and technology professionals.

(iii) Technical Neutral Procedures. The Technical Neutral shall evaluate whether the Derived Data described in the applicable Transformation Certificate satisfies the objective criteria set forth in Section 7.5(a), applying commercially reasonable standards and considering the totality of the relevant technical evidence. Each party shall have the opportunity to submit written submissions and supporting technical evidence to the Technical Neutral, and the Technical Neutral may, in his or her discretion, conduct interviews or demonstrations with technical representatives of either party. The Technical Neutral shall render a written determination within forty-five (45) calendar days of his or her appointment (or such longer period as the parties may agree in writing), setting forth the basis for the determination. The Technical Neutral's determination shall be final and binding on both parties, not subject to appeal, and enforceable in any court of competent jurisdiction.

(iv) Costs. Each party shall bear its own internal costs and legal fees incurred in connection with the Technical Neutral process. The fees and expenses of the Technical Neutral shall be shared equally by the parties.

(v) Interim Rights. Pending resolution of a Transformation Challenge through the procedures set forth in this Section 7.5(d), Licensee may continue to use the disputed Derived Data for all internal purposes under Section 7.3 without restriction. However, Licensee shall not commence third-party commercial distribution of the specific category of Derived Data that is the subject of the pending Transformation Challenge until the dispute is resolved. For the avoidance of doubt, a pending Transformation Challenge with respect to one category of Derived Data shall not affect, restrict, or delay Licensee's right to commercialize other categories of Derived Data (including QuorumScore™ outputs, which are pre-certified under Section 7.4(c) and are not subject to the Transformation Challenge process).

Section 7.6 — Intellectual Property Protections for Licensee's Proprietary Technology

(a) QuorumScore™ and Proprietary Models. Licensor acknowledges that Licensee's proprietary models and algorithms (including, without limitation, QuorumScore™, United States Patent and Trademark Office application Serial No. 97/482,116) constitute Licensee's valuable trade secrets and confidential information, and that Licensee has invested, and continues to invest, substantial resources in the development, refinement, and maintenance of such models and algorithms. Licensor agrees as follows:

(i) Licensor shall not assert, and hereby irrevocably waives, any intellectual property claim, ownership interest, license right, or other right or interest with respect to QuorumScore™, its model architecture, trained parameters, model weights, source code, object code, or any outputs thereof, whether arising under copyright, patent, trade secret, database rights, or any other theory of intellectual property law.

(ii) Licensor shall not reverse-engineer, decompile, disassemble, or otherwise attempt to derive the algorithms, model architecture, training methodologies, feature engineering processes, or other proprietary elements of QuorumScore™ or any other Licensee proprietary model from any data, outputs, reports, Transformation Certificates, or other information provided or made available by Licensee to Licensor under or in connection with this Agreement.

(iii) Licensor shall not use any information obtained from Transformation Certificates, audit processes, or other interactions under this Article VII to develop, train, enhance, or refine Licensor's own competing models, products, or services.

(b) Patent and Trade Secret Reservation. Nothing in this Agreement shall be construed to grant Licensor any rights under any patent, patent application (including provisional applications), trade secret, copyright, trademark, service mark, or other intellectual property right of Licensee, whether existing as of the Effective Date or hereafter acquired, developed, or registered. All intellectual property rights of Licensee are expressly reserved.

[Drafting Note: Quorum's IP portfolio is managed by Caldwell & Strauss LLP (Angela Firth, Partner), Quorum's regular outside intellectual property counsel. We have coordinated with Angela Firth to confirm that the provisions of this Section 7.6 are consistent with Quorum's existing IP portfolio protections, including the pending QuorumScore™ trademark application (Serial No. 97/482,116) and two provisional patent applications related to QuorumScore™'s underlying model architecture. This cross-reference is included for internal tracking purposes and will be removed from the version transmitted to Kessler Pruitt & Ashe.]

Section 7.7 — Post-Termination Treatment of Derived Data

(a) Survival of Ownership. Licensee's ownership of all Derived Data (including all intellectual property rights therein) shall survive expiration or termination of this Agreement for any reason, whether by mutual agreement, material breach, insolvency, convenience, or otherwise. Expiration or termination of this Agreement shall not divest Licensee of any ownership rights in Derived Data that existed prior to the effective date of expiration or termination.

(b) Post-Termination Use of Derived Data. Following expiration or termination of this Agreement, Licensee shall have the perpetual, irrevocable, worldwide right to continue to use, retain, store, copy, modify, adapt, enhance, analyze, and commercialize all Derived Data (including Sufficiently Transformed Outputs and Model Inference Outputs) existing as of the effective date of expiration or termination, without limitation and without any further obligation to Licensor, provided that Licensee continues to comply with Section 7.4(d) (restrictions on raw data distribution) with respect to any Derived Data commercialized to third parties following termination.

Licensee acknowledges that its right to create new Derived Data using Licensed Data terminates upon the effective date of expiration or termination of the license grant under Article III. For the avoidance of doubt, Licensee's right to create further derivatives of existing Derived Data (e.g., by combining existing Derived Data with Licensee's First-Party Data or by applying new analytical methodologies to existing Derived Data) shall not be affected by the termination of the license grant, provided that Licensee does not access or use raw Licensed Data in such further derivation activities.

(c) Return/Destruction of Licensed Data. As set forth in Article XII (Termination), upon expiration or termination of this Agreement, Licensee shall return or destroy all Licensed Data (i.e., raw Licensor data in Licensee's possession or control) within ninety (90) calendar days of the effective date of expiration or termination, in accordance with the procedures set forth in Section 12.4 of this Agreement. Licensee shall deliver to Licensor a written certification of return or destruction signed by a duly authorized officer of Licensee.

For the avoidance of doubt, the obligation to return or destroy Licensed Data under this Section 7.7(c) and Article XII does not extend to Derived Data, Model Inference Outputs, trained models (including model weights and parameters), or any other Licensee intellectual property. Licensee may retain all such materials indefinitely following expiration or termination.

(d) Trained Model Retention. Licensee may retain, use, and continue to deploy all trained machine-learning models, statistical models, and other computational models, including all model weights, parameters, coefficients, and configuration files, notwithstanding that such models were trained, calibrated, or validated using Licensed Data (among other inputs). Licensee shall not be required to retrain, modify, degrade, or destroy any model, or any component thereof, solely because Licensed Data was used as a training input during the development of such model. This provision shall survive expiration or termination of this Agreement in perpetuity.

[Drafting Note: The trained model retention provision is critical. Modern machine-learning models encode learned statistical relationships in their weights and parameters — they do not "contain" the underlying training data. Requiring Quorum to retrain or destroy models post-termination would be technically impractical, commercially devastating, and disproportionate to any legitimate interest of Meridian. We anticipate that Meridian will push back on this provision. Quorum should be prepared to explain the technical distinction between model weights (mathematical values) and training data (raw records) at the negotiation session. If necessary, we can engage Dr. Krishnamurthy to prepare a brief technical memorandum for Meridian's counsel.]

Section 7.8 — Representations and Covenants

(a) Licensee Representations and Covenants. Licensee represents, warrants, and covenants to Licensor as follows:

(i) Licensee's creation, use, storage, processing, and distribution of Derived Data shall comply in all material respects with all applicable federal, state, and local laws, rules, and regulations, including, without limitation, the Fair Credit Reporting Act, 15 U.S.C. § 1681 et seq. ("FCRA"); the Gramm-Leach-Bliley Act, 15 U.S.C. § 6801 et seq. ("GLBA"); the California Consumer Privacy Act of 2018, as amended by the California Privacy Rights Act of 2020, Cal. Civ. Code § 1798.100 et seq. ("CCPA/CPRA"); applicable state insurance data-security laws and regulations (including the NAIC Insurance Data Security Model Law, as adopted by applicable states); and any other data protection, privacy, or information security law applicable to Licensee's business activities.

(ii) Licensee shall implement and maintain commercially reasonable administrative, technical, and physical data-security measures in connection with the storage, processing, transmission, and distribution of Derived Data, consistent with the requirements of the Data Security Addendum attached to this Agreement as Appendix IX-A and with Licensee's obligations under applicable law.

(iii) Licensee shall not knowingly create, distribute, or commercialize any Derived Data in a manner that violates the rights of any individual consumer or business entity under applicable law, including any rights of privacy, data protection, or fair credit reporting.

(b) Licensor Representations and Covenants. Licensor represents, warrants, and covenants to Licensee as follows:

(i) Licensor has the right, power, and authority to grant the license set forth in Article III of this Agreement, and such license grant is sufficient in scope to permit Licensee to create, use, retain, and commercialize Derived Data as contemplated by this Article VII, without infringing or misappropriating any intellectual property right or other right of Licensor or any third party.

(ii) To Licensor's knowledge as of the Effective Date, the Licensed Data does not infringe, misappropriate, or otherwise violate the intellectual property rights of any third party in a manner that would materially impair Licensee's ability to create, use, or commercialize Derived Data as contemplated by this Article VII.

(iii) Licensor shall not, during the Term of this Agreement, enter into any agreement with a third party, or take any action, that would materially impair or restrict Licensee's rights under this Article VII with respect to Derived Data, Sufficiently Transformed Outputs, or Model Inference Outputs.

Section 7.9 — Audit Rights (Limited)

(a) Licensor Audit Right. Licensor shall have the right, exercisable no more than once per calendar year and upon not less than thirty (30) calendar days' prior written notice to Licensee, to engage an independent third-party auditor (mutually agreed by the parties, such agreement not to be unreasonably withheld, conditioned, or delayed) to verify that Licensee's third-party commercialization of Derived Data complies with the restrictions set forth in Section 7.4(d) (restrictions on raw data distribution). The independent auditor shall be bound by a written confidentiality agreement in form and substance reasonably acceptable to both parties prior to commencing any audit activities.

The scope of any such audit shall be limited solely to confirming Licensee's compliance with Section 7.4(d) and shall not extend to, and the auditor shall not be permitted to review, examine, access, or evaluate:

(i) Licensee's proprietary algorithms, model architectures, source code, object code, model weights, model parameters, training methodologies, or feature engineering processes;

(ii) Licensee's trade secrets or confidential information beyond what is strictly necessary to verify compliance with Section 7.4(d);

(iii) Licensee's commercial terms, pricing, revenue, or financial arrangements with third-party customers or licensees; or

(iv) any Derived Data, Model Inference Outputs, or analytical work product that is not the subject of third-party commercialization.

Any audit shall be conducted during normal business hours at Licensee's offices, with minimal disruption to Licensee's business operations, and shall be completed within a period not to exceed fifteen (15) business days.

Licensor shall bear all costs and expenses of the audit, including the fees of the independent auditor, unless the audit reveals a material violation of Section 7.4(d), in which case Licensee shall reimburse Licensor for the reasonable, documented, out-of-pocket costs of the audit (not to exceed $75,000 per audit). In the event of a material violation, Licensee shall promptly take commercially reasonable corrective measures to cure the violation and shall provide Licensor with a written report describing the corrective measures taken within thirty (30) calendar days of notification of the audit findings.

III. DRAFTING NOTES AND STRATEGIC COMMENTARY

[Note: This section contains internal Calderwood & Harkness analysis and strategic commentary for Quorum's legal team. This section will not be included in any version transmitted to Kessler Pruitt & Ashe LLP or Meridian.]

A. Summary of Concessions Made by Quorum in This Revised Framework

In developing this compromise framework, Quorum has made several meaningful concessions from its original January 2023 position. It is important that Carolyn Whitford and Gregory Stahl understand the scope of these concessions and the strategic rationale for each, so that Quorum can present the compromise credibly as a genuine movement from its prior position.

Introduction of the "Sufficiently Transformed Output" Concept and Objective Transformation Standard (Section 7.5). Quorum's original January 2023 draft contained no limitations on Quorum's right to commercialize Derived Data to third parties — Quorum could distribute any Derived Data, regardless of its proximity to the raw Licensed Data, without notice to or involvement by Meridian. The revised framework introduces the concept of "Sufficiently Transformed Output" and establishes a set of objective criteria (no extractability, no raw data inclusion, material transformation, aggregation/anonymization threshold) that Derived Data must satisfy before it can be commercialized to third parties. This is a meaningful concession because it accepts, in principle, that Quorum's raw-data distribution is prohibited and that there must be a demonstrable transformation before third-party commercialization is permitted. However, the criteria are objective and measurable — they do not depend on Meridian's subjective judgment — and the QuorumScore™ carve-out ensures that Quorum's most commercially significant product is not subject to case-by-case review.

Transformation Certificate Notification Mechanism and Right to Challenge (Sections 7.5(b) and 7.5(c)). Quorum's original draft provided for no Meridian involvement of any kind in Quorum's commercialization decisions. The revised framework introduces a Transformation Certificate notification mechanism that provides Meridian with visibility into Quorum's commercialization activities and a thirty-day challenge window. This represents a departure from Quorum's original position and provides Meridian with a procedural mechanism to raise concerns. Critically, however, the Transformation Certificate is a notification, not a consent request — Quorum's right to commercialize is not contingent on Meridian's approval or non-objection, and Meridian's failure to respond within thirty days results in deemed approval.

Limited Annual Audit Rights (Section 7.9). Quorum's original draft contained no audit rights for Meridian. The revised framework grants Meridian a limited annual audit right, exercised through an independent third-party auditor, focused solely on verifying compliance with the raw-data non-distribution restriction. This provides Meridian with a practical enforcement mechanism while protecting Quorum's proprietary information through strict scope limitations.

B. Items Rejected

Despite the foregoing concessions, the revised framework categorically rejects several of Meridian's core demands:

No License-Back of Any Kind. Meridian demanded a perpetual, irrevocable, worldwide, royalty-free, fully paid-up, sublicensable license-back on all Derived Data, which would have granted Meridian the right to use, copy, modify, and distribute Quorum's Derived Data for Meridian's own commercial purposes. This demand is categorically rejected in Section 7.2(c). The license-back demand is commercially unacceptable because it would allow Meridian to benefit from Quorum's investment without corresponding compensation, would create competitive risks (Meridian could use Quorum's Derived Data to develop competing products or to enhance its offerings to Quorum's competitors), and has no basis in market practice for data-licensing transactions.

No Revenue Share. Meridian demanded a 15% revenue share on net revenues from third-party commercialization of Derived Data. This demand is categorically rejected in Section 7.4(b). As noted above, the revenue share demand is unsupported by market benchmarks (per the Pinnacle Ridge Adv

PRIVILEGED & CONFIDENTIAL — ATTORNEY WORK PRODUCT — DRAFT FOR DISCUSSION PURPOSES ONLY
REVISED ARTICLE VII — DERIVED DATA AND INTELLECTUAL PROPERTY RIGHTS: QUORUM COMPROMISE FRAMEWORK
Proposed Revisions to Master Data License Agreement Between Quorum Insurance Holdings Corp. and Meridian Credit intelligence, Inc.
Date: June 2023
Matter No.: C&H Matter No. 1025-00008
Prepared by: Wendy Schultz, Associate, intellectual Property, Calderwood & Harkness LLP
With contributions from:
•  Omar Siddiqui, Associate, intellectual Property, Calderwood & Harkness LLP
•  Patricia Heng, Senior Counsel, Technology Transactions, Calderwood & Harkness LLP
NOTICE: This document is protected by the attorney-client privilege and the work product doctrine. It is intended solely for the use of the client, Quorum Insurance Holdings Corp., and is being circulated within Calderwood & Harkness LLP and to designated Quorum personnel for purposes of legal advice in connection with the negotiation of the Master Data License Agreement with Meridian Credit intelligence, Inc. This document should not be disclosed, distributed, or disseminated to any third party without the prior written consent of Calderwood & Harkness LLP.
Note: This document sets forth Quorum Insurance Holdings Corp.'s ("Quorum") revised proposed Article VII provisions for the Master Data License Agreement ("MDLA") with Meridian Credit intelligence, Inc. ("Meridian"), reflecting a compromise framework developed in advance of the June 30, 2023 third negotiation session in New York. The proposed provisions herein are intended to bridge the impasse over derived-data rights that has persisted since the first in-person negotiation session and to present a commercially reasonable framework that protects Quorum's core interests while acknowledging Meridian's legitimate concerns regarding its raw data assets.
Right-click to update Table of Contents
I. INTRODUCTORY MEMORANDUM / COVER NOTE
A. Purpose and Context
This revised Article VII represents Quorum's good-faith attempt to bridge the impasse over derived-data rights that has persisted since the January 18, 2023 first in-person negotiation session and remained unresolved through the March 9, 2023 second session in Atlanta. The parties have exchanged multiple marked-up drafts, engaged in extensive telephonic discussions between counsel, and have made substantial progress on many other articles of the MDLA. Article VII, however, remains the single most commercially significant open issue in the agreement, and its resolution is a prerequisite to closing the transaction on the timeline contemplated by both parties' respective executive teams.
Quorum's Original Position (January 2023 Draft of Article VII). In its initial draft circulated on January 12, 2023, Quorum proposed that Quorum own all intellectual property in Derived Data with unrestricted rights to use, sublicense, and commercialize such Derived Data without additional fees, royalty obligations, or Meridian consent. Under Quorum's original formulation, Derived Data was treated as Quorum's proprietary work product from the moment of creation, and Meridian had no continuing interest in, or rights with respect to, any analytical outputs generated by Quorum's processing of the Licensed Data. This position reflected Quorum's view — which we continue to believe is correct as a matter of law — that the analytical outputs produced by Quorum's proprietary algorithms, combined with Quorum's First-Party Data, represent independent works of authorship and invention that do not infringe or implicate Meridian's intellectual property rights.
Meridian's Counter-Position (November 2022 Form MDLA and Subsequent Exchanges). Meridian, through its outside counsel at Kessler Pruitt & Ashe LLP (Graham Whitley, Nina Chastain, and Derek Fontaine), advanced a dramatically different framework in the November 4, 2022 form MDLA and in subsequent exchanges. Meridian's position can be summarized as follows:
(a) Derived Data constitutes a "derivative work" under the Copyright Act (17 U.S.C. § 101 et seq.) and is therefore subject to Meridian's underlying intellectual property rights in the Licensed Data;
(b) Meridian demands a perpetual, royalty-free license-back on all Derived Data, granting Meridian the right to use, copy, modify, and distribute Quorum's Derived Data for Meridian's own commercial purposes;
(c) Quorum would be prohibited from any third-party commercialization of Derived Data unless Meridian determines, in its sole discretion, that the Derived Data reflects "sufficient transformation" of the Licensed Data to warrant independent treatment; and
(d) On those occasions where Meridian approves third-party commercialization, Meridian demands a 15% revenue share on net revenues attributable to the commercialized Derived Data.
Quorum views each of these demands as commercially unreasonable and legally unsupported. The license-back demand would effectively allow Meridian to free-ride on Quorum's multi-million-dollar investment in data science, machine learning infrastructure, and proprietary algorithm development. The sole-discretion transformation standard would give Meridian an effective veto over Quorum's commercial strategy. And the 15% revenue share is untethered to any industry benchmark or reasonable allocation of value, particularly given that Quorum's Derived Data reflects the combined contribution of Quorum's First-Party Data, Quorum's proprietary technology, and the Licensed Data.
This Compromise Framework. This revised Article VII moves Quorum from its maximalist position while firmly rejecting Meridian's core demands for a license-back, sole-discretion transformation standard, and revenue share. The key principles of the compromise are as follows:
(1) Quorum Retains Full Ownership of Derived Data. As between the parties, Quorum shall own all right, title, and interest in and to all Derived Data, consistent with the well-established principle that licensees own the work product they create using licensed inputs.
(2) No License-Back to Meridian. Quorum does not grant, and shall not be deemed to have granted, any license or right of access to Meridian in or to Quorum's Derived Data, Model Inference Outputs, or proprietary models and algorithms.
(3) No Revenue Share to Meridian. Quorum shall not owe Meridian any royalty, revenue share, profit participation, or other compensation in connection with Quorum's commercialization of Derived Data.
(4) Objective Transformation Standard for Third-Party Commercialization. Quorum may commercialize Derived Data to third parties provided the outputs meet an objective, jointly agreed de-identification and transformation standard — replacing Meridian's demanded sole-discretion standard with a framework based on measurable criteria and an independent Technical Neutral dispute resolution mechanism.
(5) Specific Protection for QuorumScore™. Outputs of Quorum's proprietary QuorumScore™ predictive underwriting model are pre-certified as Sufficiently Transformed Outputs, exempting them from case-by-case review and protecting Quorum's planned B2B licensing revenue line.
We believe this framework represents a commercially reasonable middle ground that protects Meridian's legitimate interest in preventing redistribution of its raw data while recognizing Quorum's substantial independent investment in analytics and proprietary technology. The framework is consistent with market practice in comparable data-licensing transactions and provides Meridian with meaningful protections — including an objective transformation standard, a notification and challenge mechanism, and limited audit rights — that Quorum's original draft did not include.
B. Relationship to Other Negotiation Documents
This revised Article VII should be read in conjunction with the following related documents and workstreams:
Quorum's Original Proposed Article VII (January 2023). The January 12, 2023 draft of Article VII, as circulated to Kessler Pruitt & Ashe on that date, reflects Quorum's initial position. The present document supersedes that draft in its entirety and should be treated as the operative Quorum proposal for the upcoming June 30, 2023 negotiation session.
Draft Derived Data Commercialization Protocol (Appendix VII-A). A companion document, the Derived Data Commercialization Protocol, is being developed concurrently by Wendy Schultz and Omar Siddiqui as a practical operational supplement to the proposed Article VII. The Protocol will set forth the specific format and content requirements for Transformation Certificates, detailed procedures for the Technical Neutral selection and review process, and illustrative examples of outputs that would and would not qualify as Sufficiently Transformed Outputs. The Protocol is intended to be appended to the MDLA as Appendix VII-A but is designed to be amendable by mutual written agreement of the parties without a formal MDLA amendment, providing operational flexibility as the parties' practices evolve.
Pinnacle Ridge Advisory LLC Pricing Benchmark Study. The pricing benchmark study delivered by Pinnacle Ridge Advisory LLC (Franklin Osei, Managing Director) on February 14, 2023 informs the commercial rationale underlying this compromise framework. The Pinnacle Ridge study surveyed derived-data rights provisions in 23 comparable data-licensing transactions across the financial services and insurance sectors completed between 2019 and 2022. The study's principal finding — that licensee ownership of derived data is the prevailing market norm, with revenue-share provisions appearing in fewer than 15% of surveyed transactions — supports Quorum's position that ownership without revenue sharing is the commercially reasonable baseline.
Most Favored Nation Pricing Issue. The most favored nation ("MFN") pricing issue remains separately unresolved and is not addressed in this document. This document addresses only derived-data rights under Article VII. We note, however, that progress on the derived-data issue may create goodwill and momentum for resolution of the MFN pricing dispute in subsequent sessions.
C. Instruction to Counsel
This draft is formatted as proposed contract language for Article VII of the MDLA, with drafting notes and commentary included in bracketed italicized text for internal Calderwood & Harkness and Quorum review purposes. All such drafting notes and internal commentary will be removed prior to transmission of the proposed Article VII to Kessler Pruitt & Ashe LLP.
This document is intended for discussion with Carolyn Whitford, General Counsel of Quorum Insurance Holdings Corp., and Gregory Stahl, Vice President, Legal Affairs, of Quorum Insurance Holdings Corp., prior to any transmission to opposing counsel. Upon approval by Ms. Whitford and Mr. Stahl, the proposed Article VII (stripped of internal drafting notes) will be transmitted to Kessler Pruitt & Ashe LLP (Graham Whitley, Nina Chastain, and Derek Fontaine) in advance of the June 30, 2023 negotiation session, with a cover letter from Patricia Heng framing the proposal as a compromise framework and requesting that Kessler Pruitt & Ashe review and respond with a marked-up version or counter-proposal no later than June 23, 2023.
All internal Calderwood & Harkness team members should direct questions regarding this draft to Wendy Schultz (ext. 4218) or Patricia Heng (ext. 3802).
II. PROPOSED ARTICLE VII: DERIVED DATA AND INTELLECTUAL PROPERTY RIGHTS
[Note: The following constitutes the proposed contract language for Article VII of the MDLA. Bracketed italicized text represents internal drafting notes and commentary for Calderwood & Harkness and Quorum review only and will be removed from the version transmitted to Kessler Pruitt & Ashe LLP.]
ARTICLE VII
DERIVED DATA AND INTELLECTUAL PROPERTY RIGHTS
Section 7.1 — Definitions
For purposes of this Article VII, the following capitalized terms shall have the meanings set forth below. Capitalized terms used but not defined in this Article VII shall have the meanings ascribed to them elsewhere in this Agreement.
(a) "Derived Data" means any and all data, datasets, databases, indices, scores, risk assessments, models, algorithms, model weights, model parameters, statistical outputs, analytics results, reports, blended risk scores, propensity models, enriched datasets, and other works or materials created, generated, compiled, developed, or produced by or on behalf of Licensee or its Affiliates using, incorporating, referencing, or processing the Licensed Data (as defined in Article II) as an input, whether in whole or in part, and whether alone or in combination with Licensee's own First-Party Data or third-party data. For the avoidance of doubt, "Derived Data" includes, without limitation:
(i) blended risk scores that combine Licensed Data (including Meridian credit data elements) with Licensee's First-Party Data, including but not limited to Licensee's proprietary claims data, policy data, and telematics data;
(ii) propensity models developed using Licensed Data as a training input, calibration input, or validation input;
(iii) enriched datasets created by appending, merging, or integrating Licensed Data with Licensee's First-Party Data or with data from other third-party sources;
(iv) outputs of machine-learning models, artificial intelligence systems, neural networks, gradient-boosted decision trees, ensemble models, and other statistical or computational models that were trained, calibrated, validated, or otherwise developed using Licensed Data as an input, whether in whole or in part; and
(v) outputs of QuorumScore™ (Licensee's proprietary predictive underwriting model, United States Patent and Trademark Office application pending, Serial No. 97/482,116), including all scores, risk assessments, classification outputs, and predictive analytics outputs generated by QuorumScore™.
(b) "First-Party Data" means data owned, collected, generated, or licensed by Licensee independently of the Licensed Data, including, without limitation, Licensee's proprietary claims data, policyholder data, telematics data, loss-run data, agent and broker data, actuarial data, premium data, and any data acquired by Licensee from sources other than Licensor. For the avoidance of doubt, First-Party Data does not include any data derived from or incorporating the Licensed Data.
(c) "Licensed Data" shall have the meaning ascribed to such term in Section 2.1 of this Agreement, which encompasses, without limitation, the following Licensor data products and datasets:
(i) MCI Consumer Credit File™, consisting of approximately 310 million individual consumer credit records maintained by Licensor;
(ii) MCI CommercialView™, consisting of approximately 42 million commercial business credit records maintained by Licensor; and
(iii) MCI AltSignal™, consisting of approximately 85 million individual consumer alternative-data records maintained by Licensor, including utility payment data, rental payment data, telecommunications payment data, and other non-traditional credit indicators.
The specific scope, format, delivery schedule, and update frequency for the Licensed Data are set forth in Article II and the Data Delivery Specifications attached as Appendix II-A.
(d) "Sufficiently Transformed Output" means Derived Data that has been de-identified and transformed such that:
(i) no individual record of Licensed Data can be extracted, reconstructed, reverse-engineered, or re-identified from the Derived Data through commercially reasonable means;
(ii) the Derived Data does not contain, display, or transmit any raw Licensed Data fields, tradeline-level data, or consumer-level or business-level records attributable to the Licensed Data; and
(iii) the Derived Data represents a materially new and independent work reflecting Licensee's proprietary analytical methodologies, algorithms, and/or First-Party Data contributions.
The determination of whether particular Derived Data constitutes a "Sufficiently Transformed Output" shall be made solely by reference to the objective criteria set forth in Section 7.5(a) and the procedures set forth in the Derived Data Commercialization Protocol (Appendix VII-A), and shall not be subject to either party's sole or unilateral discretion.
[Drafting Note: This definition, and in particular the proviso regarding objective criteria rather than sole discretion, is the central innovation of the compromise framework. Meridian's November 2022 form MDLA proposed that Meridian would determine "sufficient transformation" in its sole discretion — a standard that would effectively give Meridian veto power over Quorum's commercial strategy. By replacing Meridian's sole-discretion standard with objective, measurable criteria and an independent Technical Neutral dispute resolution process, we preserve Quorum's autonomy while providing Meridian with a meaningful and transparent framework for protecting its raw data.]
(e) "Model Inference Outputs" means predictions, scores, classifications, probability estimates, risk assessments, or other outputs generated by a trained machine-learning model, statistical model, or other computational model during live inference (i.e., the application of the trained model to new input data for purposes of generating predictions or other outputs), where the model was trained, calibrated, or validated using Licensed Data (among other inputs) but the live inference inputs do not include Licensed Data. For the avoidance of doubt, Model Inference Outputs include outputs generated by models that have "learned" statistical patterns, relationships, or correlations from Licensed Data during the training process but that do not retain, store, or transmit the underlying Licensed Data records used during training.
[Drafting Note: This definition is critical to resolving the technical dispute that arose during the March 9 session regarding whether ML model outputs constitute "derived data." Kessler Pruitt & Ashe's position — advanced by Derek Fontaine in his March 15 letter — is that any output from a model trained on Meridian data is inherently "derived" from that data and therefore subject to Meridian's derived-data restrictions. Quorum's position, which is technically and legally correct, is that Model Inference Outputs generated from non-Meridian inputs are independent outputs. During live inference, the model applies learned statistical relationships (encoded in model weights and parameters) to new data; the underlying training data is not accessed, referenced, or transmitted. This is analogous to a human analyst who has studied credit data and then applies learned expertise to new, independent data — the analyst's conclusions are not "derived from" the training data in any legally cognizable sense. We should be prepared to support this position with expert testimony from Dr. Ananya Krishnamurthy (MIT CSAIL) if Meridian escalates this issue.]
Section 7.2 — Ownership of Derived Data
(a) Licensee Ownership. As between the parties, Licensee shall own all right, title, and interest (including all intellectual property rights, whether arising under copyright, patent, trade secret, database rights, or other applicable law) in and to all Derived Data created, generated, compiled, developed, or produced by or on behalf of Licensee or its Affiliates during the Term or at any time thereafter.
Licensor acknowledges and agrees that Derived Data does not constitute a "derivative work" (as that term is defined in 17 U.S.C. § 101 et seq. and as interpreted by applicable case law) of the Licensed Data, and that Licensor retains no intellectual property rights in or to Derived Data by virtue of Licensor's ownership of or intellectual property rights in the Licensed Data. The parties acknowledge that the creation of Derived Data involves substantial independent intellectual contribution by Licensee — including the application of Licensee's proprietary algorithms, methodologies, trade secrets, and First-Party Data — and that the resulting Derived Data constitutes an original and independent work of Licensee.
To the extent that any Derived Data is deemed under applicable law, by any court of competent jurisdiction or other governmental authority, to constitute a derivative work of the Licensed Data or otherwise to be subject to Licensor's intellectual property rights, Licensor hereby irrevocably assigns, transfers, and conveys to Licensee all of Licensor's right, title, and interest in and to such Derived Data, and agrees to execute, at Licensee's request and expense, such further documents and instruments as may be reasonably necessary to evidence, perfect, and record such assignment.
(b) Reservation of Rights in Licensed Data. Nothing in this Article VII shall be construed to transfer to Licensee any ownership interest in the Licensed Data itself. Licensor retains all right, title, and interest in and to the Licensed Data, including all intellectual property rights therein, as set forth in Article II of this Agreement. Licensee's rights to access, use, copy, process, and otherwise exploit the Licensed Data are governed solely by the license grant set forth in Article III of this Agreement and are subject to all restrictions, limitations, and conditions set forth therein and elsewhere in this Agreement.
(c) No License-Back. Licensee does not grant, and shall not be deemed to have granted, to Licensor or any Licensor Affiliate any license, sublicense, right of access, right of use, or other right or interest in or to:
(i) any Derived Data;
(ii) any Model Inference Outputs;
(iii) any of Licensee's proprietary models, algorithms, model architectures, model weights, model parameters, source code, object code, training methodologies, or data-processing pipelines;
(iv) QuorumScore™ or any outputs thereof; or
(v) any other intellectual property, trade secrets, confidential information, or proprietary materials of Licensee or its Affiliates.
Nothing in this Agreement shall be interpreted as an implied license, license by estoppel, or other grant of rights from Licensee to Licensor in any of the foregoing.
[Drafting Note: This provision directly and categorically rejects Meridian's demand, set forth in Section 7.3(b) of Meridian's November 2022 form MDLA, for a "perpetual, irrevocable, worldwide, royalty-free, fully paid-up, sublicensable license" to all Derived Data. The license-back demand is commercially unacceptable because it would effectively grant Meridian the benefit of Quorum's multi-million-dollar investment in data science without any corresponding compensation or investment by Meridian. We anticipate strong pushback from Kessler Pruitt & Ashe on this point and should prepare Carolyn Whitford for the discussion.]
(d) Model Inference Outputs. All Model Inference Outputs shall be owned exclusively by Licensee, free and clear of any claims, rights, encumbrances, or interests of Licensor. Model Inference Outputs shall not constitute Derived Data for purposes of this Agreement and shall not be subject to any restrictions, limitations, obligations, or conditions under this Article VII or elsewhere in this Agreement. Without limiting the generality of the foregoing:
(i) Model Inference Outputs generated using non-Licensor inputs shall not be subject to the commercialization provisions of Section 7.4 or the transformation standard and certification process of Section 7.5;
(ii) Licensee may use, license, sublicense, sell, distribute, and otherwise exploit Model Inference Outputs for any purpose, including third-party commercialization, without notice to, consent of, or payment to Licensor; and
(iii) Licensee's right to use and exploit Model Inference Outputs shall survive expiration or termination of this Agreement in perpetuity.
[Drafting Note: This provision resolves the technical dispute over ML model outputs squarely in Quorum's favor. During live inference, a trained model applies learned statistical patterns to new input data; the training data (including Licensed Data) is not accessed, referenced, or retransmitted. The model weights and parameters are mathematical representations of learned correlations — they do not "contain" the underlying training data in any meaningful sense. This is consistent with the emerging consensus in the data-licensing industry and is supported by the technical analysis prepared by Quorum's Chief Data Officer, Dr. Sanjay Mehrotra. If Meridian escalates this issue, we should consider engaging an independent expert to provide a technical opinion supporting this position.]
Section 7.3 — Licensee's Internal Use Rights
(a) Unrestricted Internal Use. Licensee shall have the unrestricted, perpetual (subject to the Term limitations on creating new Derived Data set forth in Section 7.7(b)), irrevocable, fully paid-up, worldwide right to use, copy, store, modify, adapt, enhance, analyze, reprocess, reformat, and create further derivatives of Derived Data for any and all internal business purposes. Without limiting the generality of the foregoing, such internal business purposes include:
(i) automated underwriting and risk assessment across Licensee's property and casualty, life, specialty, and reinsurance lines of business;
(ii) claims-fraud detection, prevention, and investigation, including the development and deployment of machine-learning models for real-time fraud scoring;
(iii) customer propensity scoring, lifetime value modeling, cross-sell and upsell targeting, and customer segmentation for marketing and distribution purposes;
(iv) actuarial analysis, including loss reserving, rate-making, experience studies, and catastrophe modeling;
(v) regulatory reporting, compliance monitoring, and responses to inquiries or examinations by state insurance departments, the National Association of Insurance Commissioners, or other regulatory authorities;
(vi) internal research and development, including the development, training, testing, validation, and refinement of new and existing predictive models, algorithms, and analytical tools;
(vii) enterprise risk management, including portfolio analysis, concentration risk assessment, and reinsurance purchasing decisions; and
(viii) any other lawful internal business purpose as determined by Licensee in its sole discretion.
(b) Affiliate Use. Licensee may make Derived Data (including Sufficiently Transformed Outputs and Model Inference Outputs) available to its Affiliates for internal business purposes on the same terms and subject to the same conditions as apply to Licensee under this Article VII, without additional fee, royalty, or compensation to Licensor and without any requirement for prior notice to or consent from Licensor. For purposes of this Section 7.3(b), "Affiliates" shall have the meaning ascribed to such term in Section 1.1 of this Agreement, and each Affiliate receiving Derived Data shall be bound by the confidentiality obligations and data security requirements applicable to Licensee under this Agreement.
(c) QuorumScore™ Internal Use. Without limiting the generality of Sections 7.3(a) and 7.3(b), Licensee's use of QuorumScore™ (USPTO application Serial No. 97/482,116) and all outputs thereof for Licensee's internal underwriting, pricing, risk-management, and other internal business purposes shall be unrestricted and shall not require any notice to, consent of, or payment to Licensor, regardless of the extent to which Licensed Data was used as a training input, calibration input, or validation input in the development or refinement of QuorumScore™.
Section 7.4 — Commercialization of Derived Data to Third Parties
(a) Permitted Third-Party Commercialization. Subject to the conditions set forth in this Section 7.4 and Section 7.5, Licensee shall have the right to license, sublicense, sell, distribute, publish, or otherwise make available Derived Data to third parties (including entities outside Licensee's corporate group) for commercial purposes, provided that such Derived Data constitutes a "Sufficiently Transformed Output" as defined in Section 7.1(d) and as determined pursuant to the objective criteria and procedures set forth in Section 7.5 and the Derived Data Commercialization Protocol attached hereto as Appendix VII-A.
For the avoidance of doubt, Licensee's right to commercialize Derived Data under this Section 7.4 extends to any lawful commercial use, including, without limitation, licensing of scores, risk assessments, and analytical outputs to insurers, reinsurers, financial institutions, healthcare organizations, government agencies, and other entities, whether on a subscription basis, per-query basis, or any other pricing model determined by Licensee in its sole discretion.
(b) No Revenue Share. Licensee shall not owe Licensor any royalty, revenue share, profit participation, milestone payment, or other compensation of any kind in connection with Licensee's commercialization, licensing, sale, distribution, or other exploitation of Derived Data, Sufficiently Transformed Outputs, or Model Inference Outputs to third parties. All revenues, profits, and other economic benefits arising from such commercialization shall belong exclusively to Licensee.
[Drafting Note: This provision directly and categorically rejects Meridian's demand, set forth in Section 7.5(d) of Meridian's November 2022 form MDLA, for a 15% revenue share on net revenues attributable to third-party commercialization of Derived Data. The revenue share demand is commercially unreasonable for several reasons: (i) it is not supported by market practice, as confirmed by the Pinnacle Ridge Advisory study; (ii) it fails to account for Quorum's substantial independent investment in technology, data science talent, and First-Party Data; (iii) it would create perverse incentives for Meridian to inflate the scope of "Derived Data" to maximize its revenue share; and (iv) it is inconsistent with the parties' agreed license fee structure under Article VI, which already compensates Meridian for Quorum's use of the Licensed Data. We should be prepared to articulate these points clearly at the June 30 session.]
(c) QuorumScore™ Commercialization Carve-Out. Outputs of QuorumScore™ (including, without limitation, all scores, risk assessments, classifications, probability estimates, and predictive analytics outputs generated by QuorumScore™) shall be deemed Sufficiently Transformed Outputs for all purposes of this Section 7.4 and this Article VII, without further review, certification, or determination under Section 7.5, provided that such outputs do not contain or transmit any raw Licensed Data fields, tradeline-level data, or consumer-level or business-level records directly attributable to the Licensed Data.
For the avoidance of doubt, QuorumScore™ outputs that are numerical scores, probabilistic assessments, risk classifications, or similar analytical outputs — as distinguished from raw or lightly processed Licensed Data — shall be presumed to satisfy the foregoing condition, and Licensor shall bear the burden of demonstrating that any particular QuorumScore™ output fails to satisfy such condition.
[Drafting Note: The QuorumScore™ carve-out is essential to protect Quorum's competitive position and its planned B2B licensing revenue line. QuorumScore™ (USPTO application Serial No. 97/482,116) represents Quorum's single most significant proprietary technology asset, blending Meridian credit data with Quorum's own first-party claims, policy, and telematics data through proprietary machine-learning algorithms (including gradient-boosted decision trees and deep neural network ensembles). The resulting scores represent Quorum's independent intellectual contribution and are central to Quorum's planned B2B licensing business, which is projected to generate $40–50 million in annual revenue by 2027 based on internal forecasts reviewed by Quorum's Board of Directors. Subjecting QuorumScore™ outputs to a case-by-case Transformation Certificate process would create unacceptable commercial uncertainty and could delay or impede Quorum's go-to-market strategy. This carve-out should be treated as a non-negotiable element of the compromise framework; Quorum should be prepared to walk away from the MDLA if Meridian insists on subjecting QuorumScore™ to case-by-case review or revenue sharing.]
(d) Restrictions on Raw Data Distribution. For the avoidance of doubt, nothing in this Article VII permits Licensee to distribute, sublicense, sell, or otherwise make available to any third party:
(i) the Licensed Data in raw or unprocessed form;
(ii) any dataset that contains raw Licensed Data fields, tradeline-level data, or consumer-level or business-level records directly attributable to the Licensed Data, including, without limitation, individual credit scores generated by Licensor, individual payment histories, account balances, credit limits, delinquency statuses, or other data elements sourced from MCI Consumer Credit File™, MCI CommercialView™, or MCI AltSignal™ in an identifiable or record-level form; or
(iii) any output, report, dataset, or other material from which raw Licensed Data records can be extracted, reconstructed, or reverse-engineered through commercially reasonable technical means, including, without limitation, through the application of record linkage techniques, statistical disclosure attacks, or other re-identification methodologies that are commercially available or widely known in the data science community as of the date of distribution.
Any distribution of Licensed Data to third parties is governed exclusively by the terms and restrictions of Article III and, if applicable, Article IV (Permitted Disclosures), and nothing in this Article VII expands or modifies Licensee's rights under those Articles.
Section 7.5 — Objective Transformation Standard and Procedures
(a) Objective Criteria. Whether Derived Data constitutes a "Sufficiently Transformed Output" for purposes of Section 7.4 shall be determined solely by reference to the following objective criteria, and not by either party's sole, unilateral, or unfettered discretion:
(i) No Extractability. Individual consumer records or business records from the Licensed Data cannot be extracted, reconstructed, or re-identified from the Derived Data through commercially reasonable technical means. For purposes of this criterion, "commercially reasonable technical means" means technical methods, tools, and approaches that are commercially available, widely known in the data science and privacy engineering communities, and economically feasible to deploy, and does not include theoretical attacks requiring extraordinary computational resources, access to non-public information, or expertise materially beyond the state of the art.
(ii) No Raw Data Inclusion. The Derived Data does not contain, display, transmit, or make accessible any raw Licensed Data fields, including, without limitation, individual tradeline data, payment histories, account balances, credit limits, delinquency statuses, inquiry records, or identifying information (such as names, Social Security numbers, dates of birth, or addresses) sourced from MCI Consumer Credit File™, MCI CommercialView™, or MCI AltSignal™.
(iii) Material Transformation. The Derived Data reflects material analytical, statistical, algorithmic, or computational transformation by Licensee, incorporating Licensee's proprietary methodologies, algorithms, trained model parameters, and/or First-Party Data, such that the Derived Data constitutes a materially new and independent work that is qualitatively and substantively different from the Licensed Data used as an input. The mere reformatting, reorganizing, sorting, filtering, or subsetting of Licensed Data, without material analytical or computational transformation, shall not constitute a Sufficiently Transformed Output.
(iv) Aggregation/Anonymization Threshold. To the extent the Derived Data includes aggregate or statistical outputs (such as summary statistics, cohort-level risk assessments, geographic risk indices, or industry trend analyses), such outputs shall be derived from a minimum cohort of fifty (50) individual records and shall not permit the identification, whether directly or through linkage with other reasonably available information, of any individual consumer or business entity.
[Drafting Note: We have proposed a minimum cohort of 50 records for the aggregation/anonymization threshold. This number is consistent with the HIPAA Safe Harbor de-identification standard (which uses a minimum cell size approach) and with common practice in the statistical disclosure control literature. Meridian may push for a higher threshold (e.g., 100 or 250 records). We believe 50 is defensible and commercially reasonable. The specific number is bracketed as a potential negotiation point, but we should not go below 50.]
(b) Certification Process. Prior to the first commercial distribution to a third party of any new category of Derived Data (other than QuorumScore™ outputs, which are pre-certified as Sufficiently Transformed Outputs under Section 7.4(c) and are not subject to this Section 7.5(b)), Licensee shall deliver to Licensor a written certification (a "Transformation Certificate") that includes the following:
(i) a general description of the category of Derived Data to be commercialized, including the nature of the analytical outputs, the intended use cases, and the types of third-party recipients;
(ii) a written confirmation, signed by a duly authorized officer of Licensee, that the Derived Data meets each of the objective criteria set forth in Section 7.5(a)(i) through (iv); and
(iii) a description of the transformation methodology applied to the Licensed Data at a level of detail sufficient to demonstrate compliance with the criteria in Section 7.5(a), provided that Licensee shall not be required to disclose its proprietary algorithms, model architectures, source code, trade secrets, or other confidential information beyond the level of detail reasonably necessary to demonstrate compliance.
For the avoidance of doubt, the Transformation Certificate is a notification mechanism. Licensee's right to commercialize Sufficiently Transformed Outputs is not contingent upon Licensor's approval, consent, acknowledgment, or non-objection. The delivery of a Transformation Certificate is a condition to Licensee's right to commence commercial distribution of the applicable category of Derived Data, but Licensor's failure to respond to or acknowledge receipt of a Transformation Certificate shall not impair Licensee's rights hereunder.
(c) Licensor's Right to Challenge. Licensor may, within thirty (30) calendar days following Licensor's receipt of a Transformation Certificate (the "Challenge Period"), deliver to Licensee a written objection (a "Transformation Challenge") setting forth in reasonable detail:
(i) the specific criterion or criteria in Section 7.5(a) that Licensor believes the Derived Data does not satisfy; and
(ii) the factual and technical basis for Licensor's belief, including, where applicable, a description of the technical methodology by which Licensor contends that Licensed Data could be extracted, reconstructed, or re-identified from the Derived Data.
A Transformation Challenge that fails to identify a specific criterion in Section 7.5(a) and provide a reasonable factual and technical basis for the challenge shall be deemed insufficient and shall not constitute a valid objection under this Section 7.5(c).
If Licensor does not deliver a valid Transformation Challenge within the Challenge Period, the Derived Data described in the applicable Transformation Certificate shall be irrevocably deemed a Sufficiently Transformed Output for all purposes of this Agreement, and Licensee shall be free to commence and continue commercial distribution of such Derived Data without further notice to or involvement by Licensor.
(d) Dispute Resolution for Transformation Disputes. If Licensor delivers a valid and timely Transformation Challenge under Section 7.5(c), the following dispute resolution procedures shall apply:
(i) Good-Faith Negotiation. The parties shall first attempt to resolve the dispute through good-faith negotiation between designated representatives of each party (at the senior vice president level or above) for a period of thirty (30) calendar days following Licensee's receipt of the Transformation Challenge (the "Negotiation Period"). Each party shall make its designated representative reasonably available for meetings and calls during the Negotiation Period, and the parties shall exchange relevant technical information on a confidential basis to facilitate resolution.
(ii) Technical Neutral. If the dispute is not resolved through good-faith negotiation within the Negotiation Period, either party may, by written notice to the other party, submit the dispute to an independent technical expert (a "Technical Neutral") for binding determination. The Technical Neutral shall be a qualified data-science professional with at least fifteen (15) years of experience in data analytics, data privacy, statistical disclosure control, or a closely related field, and shall not have any material conflict of interest with either party. The parties shall use good-faith efforts to agree upon the selection of the Technical Neutral within fifteen (15) calendar days of the notice submitting the dispute. If the parties are unable to agree upon the selection of the Technical Neutral within such fifteen (15) day period, either party may request that the American Arbitration Association appoint the Technical Neutral from its panel of qualified data-science and technology professionals.
(iii) Technical Neutral Procedures. The Technical Neutral shall evaluate whether the Derived Data described in the applicable Transformation Certificate satisfies the objective criteria set forth in Section 7.5(a), applying commercially reasonable standards and considering the totality of the relevant technical evidence. Each party shall have the opportunity to submit written submissions and supporting technical evidence to the Technical Neutral, and the Technical Neutral may, in his or her discretion, conduct interviews or demonstrations with technical representatives of either party. The Technical Neutral shall render a written determination within forty-five (45) calendar days of his or her appointment (or such longer period as the parties may agree in writing), setting forth the basis for the determination. The Technical Neutral's determination shall be final and binding on both parties, not subject to appeal, and enforceable in any court of competent jurisdiction.
(iv) Costs. Each party shall bear its own internal costs and legal fees incurred in connection with the Technical Neutral process. The fees and expenses of the Technical Neutral shall be shared equally by the parties.
(v) Interim Rights. Pending resolution of a Transformation Challenge through the procedures set forth in this Section 7.5(d), Licensee may continue to use the disputed Derived Data for all internal purposes under Section 7.3 without restriction. However, Licensee shall not commence third-party commercial distribution of the specific category of Derived Data that is the subject of the pending Transformation Challenge until the dispute is resolved. For the avoidance of doubt, a pending Transformation Challenge with respect to one category of Derived Data shall not affect, restrict, or delay Licensee's right to commercialize other categories of Derived Data (including QuorumScore™ outputs, which are pre-certified under Section 7.4(c) and are not subject to the Transformation Challenge process).
Section 7.6 — Intellectual Property Protections for Licensee's Proprietary Technology
(a) QuorumScore™ and Proprietary Models. Licensor acknowledges that Licensee's proprietary models and algorithms (including, without limitation, QuorumScore™, United States Patent and Trademark Office application Serial No. 97/482,116) constitute Licensee's valuable trade secrets and confidential information, and that Licensee has invested, and continues to invest, substantial resources in the development, refinement, and maintenance of such models and algorithms. Licensor agrees as follows:
(i) Licensor shall not assert, and hereby irrevocably waives, any intellectual property claim, ownership interest, license right, or other right or interest with respect to QuorumScore™, its model architecture, trained parameters, model weights, source code, object code, or any outputs thereof, whether arising under copyright, patent, trade secret, database rights, or any other theory of intellectual property law.
(ii) Licensor shall not reverse-engineer, decompile, disassemble, or otherwise attempt to derive the algorithms, model architecture, training methodologies, feature engineering processes, or other proprietary elements of QuorumScore™ or any other Licensee proprietary model from any data, outputs, reports, Transformation Certificates, or other information provided or made available by Licensee to Licensor under or in connection with this Agreement.
(iii) Licensor shall not use any information obtained from Transformation Certificates, audit processes, or other interactions under this Article VII to develop, train, enhance, or refine Licensor's own competing models, products, or services.
(b) Patent and Trade Secret Reservation. Nothing in this Agreement shall be construed to grant Licensor any rights under any patent, patent application (including provisional applications), trade secret, copyright, trademark, service mark, or other intellectual property right of Licensee, whether existing as of the Effective Date or hereafter acquired, developed, or registered. All intellectual property rights of Licensee are expressly reserved.
[Drafting Note: Quorum's IP portfolio is managed by Caldwell & Strauss LLP (Angela Firth, Partner), Quorum's regular outside intellectual property counsel. We have coordinated with Angela Firth to confirm that the provisions of this Section 7.6 are consistent with Quorum's existing IP portfolio protections, including the pending QuorumScore™ trademark application (Serial No. 97/482,116) and two provisional patent applications related to QuorumScore™'s underlying model architecture. This cross-reference is included for internal tracking purposes and will be removed from the version transmitted to Kessler Pruitt & Ashe.]
Section 7.7 — Post-Termination Treatment of Derived Data
(a) Survival of Ownership. Licensee's ownership of all Derived Data (including all intellectual property rights therein) shall survive expiration or termination of this Agreement for any reason, whether by mutual agreement, material breach, insolvency, convenience, or otherwise. Expiration or termination of this Agreement shall not divest Licensee of any ownership rights in Derived Data that existed prior to the effective date of expiration or termination.
(b) Post-Termination Use of Derived Data. Following expiration or termination of this Agreement, Licensee shall have the perpetual, irrevocable, worldwide right to continue to use, retain, store, copy, modify, adapt, enhance, analyze, and commercialize all Derived Data (including Sufficiently Transformed Outputs and Model Inference Outputs) existing as of the effective date of expiration or termination, without limitation and without any further obligation to Licensor, provided that Licensee continues to comply with Section 7.4(d) (restrictions on raw data distribution) with respect to any Derived Data commercialized to third parties following termination.
Licensee acknowledges that its right to create new Derived Data using Licensed Data terminates upon the effective date of expiration or termination of the license grant under Article III. For the avoidance of doubt, Licensee's right to create further derivatives of existing Derived Data (e.g., by combining existing Derived Data with Licensee's First-Party Data or by applying new analytical methodologies to existing Derived Data) shall not be affected by the termination of the license grant, provided that Licensee does not access or use raw Licensed Data in such further derivation activities.
(c) Return/Destruction of Licensed Data. As set forth in Article XII (Termination), upon expiration or termination of this Agreement, Licensee shall return or destroy all Licensed Data (i.e., raw Licensor data in Licensee's possession or control) within ninety (90) calendar days of the effective date of expiration or termination, in accordance with the procedures set forth in Section 12.4 of this Agreement. Licensee shall deliver to Licensor a written certification of return or destruction signed by a duly authorized officer of Licensee.
For the avoidance of doubt, the obligation to return or destroy Licensed Data under this Section 7.7(c) and Article XII does not extend to Derived Data, Model Inference Outputs, trained models (including model weights and parameters), or any other Licensee intellectual property. Licensee may retain all such materials indefinitely following expiration or termination.
(d) Trained Model Retention. Licensee may retain, use, and continue to deploy all trained machine-learning models, statistical models, and other computational models, including all model weights, parameters, coefficients, and configuration files, notwithstanding that such models were trained, calibrated, or validated using Licensed Data (among other inputs). Licensee shall not be required to retrain, modify, degrade, or destroy any model, or any component thereof, solely because Licensed Data was used as a training input during the development of such model. This provision shall survive expiration or termination of this Agreement in perpetuity.
[Drafting Note: The trained model retention provision is critical. Modern machine-learning models encode learned statistical relationships in their weights and parameters — they do not "contain" the underlying training data. Requiring Quorum to retrain or destroy models post-termination would be technically impractical, commercially devastating, and disproportionate to any legitimate interest of Meridian. We anticipate that Meridian will push back on this provision. Quorum should be prepared to explain the technical distinction between model weights (mathematical values) and training data (raw records) at the negotiation session. If necessary, we can engage Dr. Krishnamurthy to prepare a brief technical memorandum for Meridian's counsel.]
Section 7.8 — Representations and Covenants
(a) Licensee Representations and Covenants. Licensee represents, warrants, and covenants to Licensor as follows:
(i) Licensee's creation, use, storage, processing, and distribution of Derived Data shall comply in all material respects with all applicable federal, state, and local laws, rules, and regulations, including, without limitation, the Fair Credit Reporting Act, 15 U.S.C. § 1681 et seq. ("FCRA"); the Gramm-Leach-Bliley Act, 15 U.S.C. § 6801 et seq. ("GLBA"); the California Consumer Privacy Act of 2018, as amended by the California Privacy Rights Act of 2020, Cal. Civ. Code § 1798.100 et seq. ("CCPA/CPRA"); applicable state insurance data-security laws and regulations (including the NAIC Insurance Data Security Model Law, as adopted by applicable states); and any other data protection, privacy, or information security law applicable to Licensee's business activities.
(ii) Licensee shall implement and maintain commercially reasonable administrative, technical, and physical data-security measures in connection with the storage, processing, transmission, and distribution of Derived Data, consistent with the requirements of the Data Security Addendum attached to this Agreement as Appendix IX-A and with Licensee's obligations under applicable law.
(iii) Licensee shall not knowingly create, distribute, or commercialize any Derived Data in a manner that violates the rights of any individual consumer or business entity under applicable law, including any rights of privacy, data protection, or fair credit reporting.
(b) Licensor Representations and Covenants. Licensor represents, warrants, and covenants to Licensee as follows:
(i) Licensor has the right, power, and authority to grant the license set forth in Article III of this Agreement, and such license grant is sufficient in scope to permit Licensee to create, use, retain, and commercialize Derived Data as contemplated by this Article VII, without infringing or misappropriating any intellectual property right or other right of Licensor or any third party.
(ii) To Licensor's knowledge as of the Effective Date, the Licensed Data does not infringe, misappropriate, or otherwise violate the intellectual property rights of any third party in a manner that would materially impair Licensee's ability to create, use, or commercialize Derived Data as contemplated by this Article VII.
(iii) Licensor shall not, during the Term of this Agreement, enter into any agreement with a third party, or take any action, that would materially impair or restrict Licensee's rights under this Article VII with respect to Derived Data, Sufficiently Transformed Outputs, or Model Inference Outputs.
Section 7.9 — Audit Rights (Limited)
(a) Licensor Audit Right. Licensor shall have the right, exercisable no more than once per calendar year and upon not less than thirty (30) calendar days' prior written notice to Licensee, to engage an independent third-party auditor (mutually agreed by the parties, such agreement not to be unreasonably withheld, conditioned, or delayed) to verify that Licensee's third-party commercialization of Derived Data complies with the restrictions set forth in Section 7.4(d) (restrictions on raw data distribution). The independent auditor shall be bound by a written confidentiality agreement in form and substance reasonably acceptable to both parties prior to commencing any audit activities.
The scope of any such audit shall be limited solely to confirming Licensee's compliance with Section 7.4(d) and shall not extend to, and the auditor shall not be permitted to review, examine, access, or evaluate:
(i) Licensee's proprietary algorithms, model architectures, source code, object code, model weights, model parameters, training methodologies, or feature engineering processes;
(ii) Licensee's trade secrets or confidential information beyond what is strictly necessary to verify compliance with Section 7.4(d);
(iii) Licensee's commercial terms, pricing, revenue, or financial arrangements with third-party customers or licensees; or
(iv) any Derived Data, Model Inference Outputs, or analytical work product that is not the subject of third-party commercialization.
Any audit shall be conducted during normal business hours at Licensee's offices, with minimal disruption to Licensee's business operations, and shall be completed within a period not to exceed fifteen (15) business days.
Licensor shall bear all costs and expenses of the audit, including the fees of the independent auditor, unless the audit reveals a material violation of Section 7.4(d), in which case Licensee shall reimburse Licensor for the reasonable, documented, out-of-pocket costs of the audit (not to exceed $75,000 per audit). In the event of a material violation, Licensee shall promptly take commercially reasonable corrective measures to cure the violation and shall provide Licensor with a written report describing the corrective measures taken within thirty (30) calendar days of notification of the audit findings.
III. DRAFTING NOTES AND STRATEGIC COMMENTARY
[Note: This section contains internal Calderwood & Harkness analysis and strategic commentary for Quorum's legal team. This section will not be included in any version transmitted to Kessler Pruitt & Ashe LLP or Meridian.]
A. Summary of Concessions Made by Quorum in This Revised Framework
In developing this compromise framework, Quorum has made several meaningful concessions from its original January 2023 position. It is important that Carolyn Whitford and Gregory Stahl understand the scope of these concessions and the strategic rationale for each, so that Quorum can present the compromise credibly as a genuine movement from its prior position.
Introduction of the "Sufficiently Transformed Output" Concept and Objective Transformation Standard (Section 7.5). Quorum's original January 2023 draft contained no limitations on Quorum's right to commercialize Derived Data to third parties — Quorum could distribute any Derived Data, regardless of its proximity to the raw Licensed Data, without notice to or involvement by Meridian. The revised framework introduces the concept of "Sufficiently Transformed Output" and establishes a set of objective criteria (no extractability, no raw data inclusion, material transformation, aggregation/anonymization threshold) that Derived Data must satisfy before it can be commercialized to third parties. This is a meaningful concession because it accepts, in principle, that Quorum's raw-data distribution is prohibited and that there must be a demonstrable transformation before third-party commercialization is permitted. However, the criteria are objective and measurable — they do not depend on Meridian's subjective judgment — and the QuorumScore™ carve-out ensures that Quorum's most commercially significant product is not subject to case-by-case review.
Transformation Certificate Notification Mechanism and Right to Challenge (Sections 7.5(b) and 7.5(c)). Quorum's original draft provided for no Meridian involvement of any kind in Quorum's commercialization decisions. The revised framework introduces a Transformation Certificate notification mechanism that provides Meridian with visibility into Quorum's commercialization activities and a thirty-day challenge window. This represents a departure from Quorum's original position and provides Meridian with a procedural mechanism to raise concerns. Critically, however, the Transformation Certificate is a notification, not a consent request — Quorum's right to commercialize is not contingent on Meridian's approval or non-objection, and Meridian's failure to respond within thirty days results in deemed approval.
Limited Annual Audit Rights (Section 7.9). Quorum's original draft contained no audit rights for Meridian. The revised framework grants Meridian a limited annual audit right, exercised through an independent third-party auditor, focused solely on verifying compliance with the raw-data non-distribution restriction. This provides Meridian with a practical enforcement mechanism while protecting Quorum's proprietary information through strict scope limitations.
B. Items Rejected
Despite the foregoing concessions, the revised framework categorically rejects several of Meridian's core demands:
No License-Back of Any Kind. Meridian demanded a perpetual, irrevocable, worldwide, royalty-free, fully paid-up, sublicensable license-back on all Derived Data, which would have granted Meridian the right to use, copy, modify, and distribute Quorum's Derived Data for Meridian's own commercial purposes. This demand is categorically rejected in Section 7.2(c). The license-back demand is commercially unacceptable because it would allow Meridian to benefit from Quorum's investment without corresponding compensation, would create competitive risks (Meridian could use Quorum's Derived Data to develop competing products or to enhance its offerings to Quorum's competitors), and has no basis in market practice for data-licensing transactions.
No Revenue Share. Meridian demanded a 15% revenue share on net revenues from third-party commercialization of Derived Data. This demand is categorically rejected in Section 7.4(b). As noted above, the revenue share demand is unsupported by market benchmarks (per the Pinnacle Ridge Adv

PRIVILEGED & CONFIDENTIAL — ATTORNEY WORK PRODUCT — DRAFT FOR DISCUSSION PURPOSES ONLY

I. INTRODUCTORY MEMORANDUM / COVER NOTE

II. PROPOSED ARTICLE VII: DERIVED DATA AND INTELLECTUAL PROPERTY RIGHTS

ARTICLE VII

DERIVED DATA AND INTELLECTUAL PROPERTY RIGHTS

III. DRAFTING NOTES AND STRATEGIC COMMENTARY

Record

Parties

Dates

Amounts