MEMORANDUM
PRIVILEGED AND CONFIDENTIAL — ATTORNEY-CLIENT COMMUNICATION / ATTORNEY WORK PRODUCT This memorandum is protected by the attorney-client privilege and the work product doctrine. It is intended solely for the use of the addressees and those persons specifically identified in the distribution list. Any disclosure, distribution, or copying of this memorandum or its contents to any unauthorized person is strictly prohibited.
TO: Carolyn Whitford, General Counsel & Executive Vice President, Quorum Insurance Holdings Corp. Gregory Stahl, Vice President, Legal Affairs, Quorum Insurance Holdings Corp.
FROM: Wendy Schultz, Associate, intellectual Property Practice Group, Calderwood & Harkness LLP
CC: Leonard Cross, Partner (Responsible Partner) Patricia Heng, Senior Counsel, Technology Transactions Practice Group Omar Siddiqui, Associate, intellectual Property Practice Group
DATE: June 2023
RE: QuorumScore™ intellectual Property Protection Strategy in the Context of the Master Data License Agreement Negotiation with Meridian Credit intelligence, Inc.
Matter No.: C&H Matter No. 1025-00008
I. Executive Summary
The purpose of this memorandum is to analyze the intellectual property protection strategy for Quorum Insurance Holdings Corp.'s ("Quorum" or the "Company") proprietary predictive underwriting model, QuorumScore™, in the context of the ongoing negotiation of a Master Data License Agreement ("MDLA") with Meridian Credit intelligence, Inc. ("Meridian"). This memorandum has been prepared at the request of Quorum's Office of General Counsel and should be read in conjunction with the January 2023 memorandum addressing copyright and trade secret analysis of machine-learning model outputs and the separate May 2023 memorandum critiquing Meridian's proposed "sufficient transformation" standard.
QuorumScore™ is central to Quorum's competitive differentiation in automated underwriting across its property and casualty, life, and specialty lines of business. The model represents years of investment in data science talent, proprietary algorithmic development, and the integration of Quorum's extensive first-party data assets. Beyond its internal underwriting applications, QuorumScore™ is the foundation of a planned business-to-business ("B2B") licensing revenue line that Quorum's leadership has projected at $40–50 million in annual revenue by 2027. The protection of QuorumScore™'s intellectual property is therefore not merely a defensive legal exercise but a matter of direct and material consequence to the Company's strategic growth trajectory and enterprise valuation.
Meridian's current MDLA positions — as set forth in the draft circulated by Meridian's outside counsel, Kessler Pruitt & Ashe LLP — pose significant and, in certain respects, existential threats to the integrity and commercial value of QuorumScore™. Three provisions in particular warrant urgent attention:
(a) Meridian's demand for a perpetual, royalty-free license-back on all Derived Data (as defined in the draft MDLA), which would grant Meridian unrestricted use of QuorumScore™ outputs and effectively allow Meridian to replicate or leverage Quorum's proprietary analytics for its own commercial benefit or for the benefit of Quorum's competitors;
(b) Meridian's insistence on retaining sole discretion to determine whether Derived Data has been "sufficiently transformed" to permit third-party commercialization by Quorum, which would grant Meridian an effective veto over the commercial deployment of QuorumScore™ in the B2B marketplace; and
(c) Meridian's demand for a 15% revenue share on net revenues attributable to any Derived Data that Quorum commercializes externally, which — on the projected $40–50 million in B2B revenue — would impose an annual cost of $6–7.5 million on Quorum, in addition to the $12.3 million annual license fee under the MDLA itself.
Each of these provisions independently threatens the proprietary value and commercial exploitation pathway for QuorumScore™. Taken together, they would fundamentally alter the economics and strategic rationale of the MDLA relationship and would substantially undermine the intellectual property portfolio that underpins Quorum's planned B2B licensing business.
This memorandum addresses six principal areas of analysis:
1. The current intellectual property protection posture for QuorumScore™, including patent, trade secret, trademark, and copyright protections;
2. The specific threats posed by Meridian's MDLA positions to QuorumScore™'s intellectual property and commercial value;
3. Trade secret safeguards and the risks that Meridian's proposed provisions pose to the continued viability of QuorumScore™'s trade secret protections;
4. Patent protection status and strategic recommendations to strengthen Quorum's negotiating position;
5. Recommended contractual provisions in the MDLA to insulate QuorumScore™ from Meridian's Derived Data claims and to preserve Quorum's unrestricted rights to use, sublicense, and commercialize QuorumScore™ outputs; and
6. Coordination with Caldwell & Strauss LLP (Angela Firth, partner contact), Quorum's longstanding outside intellectual property counsel, on the implications of the MDLA negotiation for Quorum's existing intellectual property portfolio.
Bottom Line: Quorum must ensure that the MDLA expressly carves out QuorumScore™ and its outputs from any Derived Data provisions that would subject QuorumScore™ to Meridian's control, license-back rights, or revenue-sharing obligations. Failure to secure this carve-out could compromise both the proprietary value and the commercial exploitation pathway for QuorumScore™, with cascading consequences for the Company's B2B licensing strategy, enterprise valuation, and competitive position in the automated underwriting marketplace. Calderwood & Harkness recommends that Quorum treat the QuorumScore™ carve-out as a non-negotiable element of the MDLA, on par with the most favored nation and pricing provisions in terms of commercial priority.
II. Background
A. QuorumScore™ — Description and Business Significance
QuorumScore™ is Quorum's proprietary predictive underwriting model designed to automate and enhance risk assessment across the Company's underwriting operations. The model operates by blending multiple data inputs from both external and internal sources to produce a suite of analytics outputs that inform and, in many cases, drive underwriting decisions. The external data inputs include credit and consumer data licensed from Meridian Credit intelligence, Inc., specifically the MCI Consumer Credit File™, MCI CommercialView™, and MCI AltSignal™ data products. These Meridian data feeds are integrated alongside Quorum's own extensive first-party data assets, which include historical claims data spanning over fifteen years of loss experience, detailed policy data across all of Quorum's lines of business, and telematics data gathered from Quorum's connected-device underwriting programs.
The model's analytical outputs include blended risk scores that synthesize credit, claims, policy, and behavioral signals into a single numeric assessment of underwriting risk; propensity models that predict the likelihood of various loss events, policy lapses, and cross-selling opportunities; and enriched datasets that combine structured and unstructured data into formats optimized for downstream analytics and decisioning. These outputs are currently deployed across Quorum's property and casualty, life, and specialty lines of business, where they have contributed to measurable improvements in loss ratios, pricing accuracy, and operational efficiency.
QuorumScore™ is the centerpiece of Quorum's underwriting competitive advantage. As the insurance industry increasingly moves toward data-driven, automated underwriting — a trend accelerated by advances in machine learning and the growing availability of alternative data sources — the ability to produce differentiated, proprietary risk analytics has become a critical source of competitive differentiation. QuorumScore™ provides Quorum with precisely this capability, and its continued development and refinement is a top strategic priority for the Company's executive leadership.
Beyond its internal applications, QuorumScore™ is the foundation for a planned B2B licensing revenue line through which Quorum intends to license QuorumScore™-derived analytics and risk scores to third-party insurers, managing general agents, and other financial services firms. Quorum's internal projections, as developed by the office of Chief Data Officer Rajesh Anand and SVP of Underwriting Analytics Diane Lowell, forecast that this B2B licensing business will generate $40–50 million in annual revenue by 2027. This revenue projection represents a material contribution to Quorum's growth strategy and has been incorporated into the Company's long-range financial plan and investor communications.
The primary internal business sponsors of QuorumScore™ are Mr. Anand and Ms. Lowell, who oversee the strategic direction and business application of the model. The technical development and data-feed integration work is led by Priya Nair, Quorum's Director of Data Engineering, whose team is responsible for the ongoing refinement of the model's feature engineering pipeline, the integration of new data sources, and the maintenance of the production-grade inference systems that generate QuorumScore™ outputs in real time.
B. Current IP Protection Posture
Quorum has pursued a multi-layered intellectual property protection strategy for QuorumScore™, encompassing patent, trade secret, trademark, and copyright protections. The current status of each layer is summarized below.
Patent Protection. QuorumScore™ is the subject of a pending patent application before the United States Patent and Trademark Office ("USPTO"), Serial No. 97/482,116. The application covers the model's architecture, including the specific neural network topology and ensemble methodology employed by QuorumScore™; the scoring methodology, including the proprietary approach to blending heterogeneous data inputs into a unified risk score; and certain algorithmic innovations developed by Quorum's data science team in the course of building and refining the model. The application was filed by Caldwell & Strauss LLP on behalf of Quorum and is currently pending — it has not yet been examined or issued. The significance of the patent application to the MDLA negotiation, and the strategic recommendations for expediting prosecution, are discussed in detail in Section V below.
Trade Secret Protection. QuorumScore™'s model weights, training data configurations, feature engineering processes, and algorithm parameters are maintained as trade secrets under Quorum's internal information security program. These elements represent the core intellectual innovation embedded in QuorumScore™ and derive independent economic value from the fact that they are not generally known to or readily ascertainable by Quorum's competitors. Access to these trade secrets is restricted to authorized personnel — principally members of Quorum's data science and underwriting analytics teams — who are subject to confidentiality obligations under their employment agreements or, in the case of contractors, under separate non-disclosure agreements. Quorum's information security program, which governs access to QuorumScore™ trade secrets, is described in further detail in Section IV.A below.
Trademark Protection. "QuorumScore" is used as a trademark of Quorum Insurance Holdings Corp. in connection with the Company's predictive underwriting analytics services. The mark is currently used with the ™ designation, indicating that Quorum claims common-law trademark rights but has not yet obtained a federal registration. We recommend that Caldwell & Strauss LLP confirm the registration status of the mark and, if a federal registration has not been filed, assess whether formal registration should be pursued. A registered trademark would provide Quorum with enhanced enforcement rights and constructive notice of ownership, both of which would be valuable in the context of the planned B2B licensing business.
Copyright Protection. The source code implementing QuorumScore™ is subject to copyright protection under the Copyright Act of 1976, 17 U.S.C. § 101 et seq., as an original literary work. Quorum owns the copyright in this source code as a work made for hire, as the code was developed by Quorum employees within the scope of their employment. Quorum has not registered the copyright with the U.S. Copyright Office, which means that, while Quorum's ownership is established, the Company would need to register the copyright before commencing an infringement action in federal court. We recommend that Quorum consider registering the copyright in the QuorumScore™ source code to preserve the availability of statutory damages and attorney's fees in the event of infringement.
Coordination with Caldwell & Strauss LLP. Angela Firth, a partner at Caldwell & Strauss LLP (100 Pearl Street, Hartford, CT 06103), is Quorum's longstanding outside intellectual property counsel. Ms. Firth and her team are advising Quorum's in-house legal team on the implications of the MDLA negotiation for Quorum's existing IP portfolio and are coordinating with Calderwood & Harkness on the derived-data IP ownership positions that are central to this memorandum. Caldwell & Strauss is not billing separately on the MDLA matter but is providing background guidance and input on patent prosecution strategy, trademark registration, and portfolio review. The recommended coordination steps are set forth in Section VIII below.
C. The MDLA Negotiation — Status of Derived-Data Provisions
The proposed MDLA between Quorum and Meridian contemplates a five-year enterprise data-licensing arrangement under which Quorum would receive access to a suite of Meridian's credit and consumer data products in exchange for aggregate fees of $60,000,000 over the initial contract term. The fee structure provides for $10,800,000 in Year 1 and $12,300,000 per year in Years 2 through 5, reflecting volume-based pricing adjustments and the phased onboarding of additional Meridian data products during the first year of the arrangement.
The MDLA negotiation, which has been conducted over the course of several months and multiple negotiation sessions involving both legal counsel and business principals, has stalled principally on two issues. The first is Meridian's refusal to grant a most favored nation ("MFN") pricing clause that would ensure Quorum receives pricing terms no less favorable than those offered to Meridian's other enterprise customers of comparable size and data consumption. The second — and the focus of this memorandum — is the deep and persistent disagreement between the parties over Quorum's rights to create, own, and commercially exploit Derived Data built on top of the licensed Meridian data.
Meridian's form MDLA, which was drafted by its outside counsel, Kessler Pruitt & Ashe LLP (Graham Whitley, lead partner), takes an expansive position on Meridian's intellectual property rights in Derived Data. Specifically, Meridian's form MDLA defines "Derived Data" broadly to encompass any data, analytics, models, scores, indices, or other outputs that "incorporate, are based on, or are traceable to" Meridian's raw data. The draft further asserts that any such Derived Data constitutes a "derivative work" under the Copyright Act and remains subject to Meridian's underlying intellectual property rights in the raw data from which it was derived. This legal theory — which we have analyzed and critiqued in prior memoranda — forms the basis for three specific demands that Meridian has advanced in the negotiation:
(a) Perpetual, Royalty-Free License-Back. Meridian demands a perpetual, irrevocable, royalty-free license to use all Derived Data created by Quorum during the term of the MDLA for Meridian's own internal and commercial purposes, without restriction.
(b) "Sufficient Transformation" Standard — Sole Discretion. Meridian proposes that Quorum be prohibited from selling, sublicensing, or otherwise commercializing Derived Data to or with third parties unless the Derived Data has been "sufficiently transformed" such that no Meridian raw data can be reverse-engineered or reconstructed from the Derived Data output. Critically, Meridian's draft reserves to Meridian the sole discretion to determine whether the "sufficient transformation" standard has been met in any given case. There are no objective criteria, thresholds, or appeal mechanisms set forth in the draft.
(c) 15% Revenue Share. In the event that Quorum receives Meridian's approval to commercialize Derived Data externally, Meridian demands a 15% revenue share on net revenues attributable to the commercialized Derived Data. The draft does not define "net revenues" with precision, nor does it provide a methodology for attributing revenues to Derived Data as distinct from other value inputs.
Quorum's position throughout the negotiation has been that Quorum should own all intellectual property in Derived Data and should have unrestricted rights to use, sublicense, and commercialize Derived Data without additional fees, revenue-sharing obligations, or the requirement of Meridian's consent. This position is grounded in the fundamental principle that Quorum's investment in data science talent, proprietary algorithms, and first-party data assets creates independent intellectual property value that is not derived from — and does not diminish — Meridian's underlying data.
At the third negotiation session, held on June 30, 2023, Quorum proposed a revised compromise framework that represents a significant step toward the middle but preserves Quorum's core interests. Under this framework: Quorum retains full ownership of all Derived Data; no license-back is granted to Meridian; Quorum may commercialize Derived Data outputs to third parties provided the outputs are sufficiently de-identified and transformed to meet an objective, jointly agreed standard (not subject to Meridian's sole discretion); no revenue share is payable to Meridian on commercialized Derived Data; and a specific carve-out is included for QuorumScore™ outputs, which would be excluded from the definition of "Derived Data" entirely and thus not subject to any of the foregoing restrictions. As of the date of this memorandum, Meridian has not formally responded to Quorum's June 2023 proposal.
It bears noting that Meridian's negotiating posture on the Derived Data provisions is influenced not only by its own commercial interests but also by the perspectives of its principal equity holders and financial advisors. Breckenridge Capital Partners Fund IV, L.P. ("Breckenridge"), which holds a 62.5% ownership stake in Meridian, has been actively involved in the MDLA negotiation through Caleb Rourke, the Breckenridge partner overseeing the Meridian investment, and Elaine Drummond, Managing Partner of Breckenridge Capital Management LLC, the fund's general partner. Additionally, Whitmore Zanthor & Co., acting through Managing Director Rachel Ng-Castillo, is advising Meridian in connection with a planned initial public offering in the 2025–2026 timeframe. These stakeholders have expressed concern that broad derived-data rights granted to Quorum (or any similarly situated enterprise customer) could erode Meridian's proprietary data moat and negatively impact the valuation narrative that Meridian intends to present to public market investors in connection with its IPO. This dynamic adds a layer of complexity to the negotiation, as Meridian's positions on Derived Data are driven in part by valuation and narrative considerations that extend beyond the bilateral commercial relationship with Quorum.
III. Threat Analysis: Impact of Meridian's MDLA Positions on QuorumScore™
The following analysis examines each of Meridian's principal MDLA positions as they specifically relate to QuorumScore™ and its outputs. The objective is to identify the precise mechanisms by which each provision threatens QuorumScore™'s intellectual property integrity, competitive value, and commercial exploitation pathway.
A. License-Back Provision
Meridian's demand for a perpetual, irrevocable, royalty-free license-back on all Derived Data presents the most immediate and direct threat to QuorumScore™'s proprietary value. If the definition of "Derived Data" is not amended to exclude QuorumScore™ outputs, this provision would grant Meridian unrestricted, perpetual use of the blended risk scores, propensity models, and enriched datasets that QuorumScore™ generates — the very outputs that represent the core of Quorum's competitive advantage in automated underwriting.
The implications of such a license-back are severe. First, Meridian would be permitted to use QuorumScore™ outputs in its own analytics and data products, effectively benefiting from Quorum's proprietary investment without compensating Quorum for that value. Second, and more concerning, Meridian could incorporate QuorumScore™ insights — including the risk-scoring methodologies and propensity models that differentiate Quorum's underwriting — into Meridian's own commercial offerings, including data products sold to Quorum's competitors. Meridian operates as a data bureau with approximately $1.42 billion in fiscal year 2023 revenue and maintains an analytics business that directly competes with the type of B2B data products Quorum plans to offer through the QuorumScore™ licensing program. The license-back would, in effect, give Meridian a royalty-free pathway to offer QuorumScore™-equivalent analytics to every insurer, reinsurer, and managing general agent in the marketplace.
Even if the license-back were nominally limited to "Meridian's own use" — a limitation that does not appear in the current draft, which grants the license without use restrictions — the practical effect would be substantially the same. Meridian's "own use" encompasses its commercial data products and analytics offerings, which are sold to the same customer base that Quorum intends to target with the B2B licensing program. The distinction between Meridian using QuorumScore™ outputs "internally" and Meridian embedding QuorumScore™ insights into products sold to Quorum's competitors is, in practice, illusory.
Impact Assessment: The license-back provision, if it captures QuorumScore™ outputs, would result in the direct erosion of QuorumScore™'s proprietary value and the potential cannibalization of Quorum's projected $40–50 million B2B licensing revenue line by 2027. This provision is, in Calderwood & Harkness's assessment, the single most damaging element of Meridian's current MDLA positions with respect to QuorumScore™.
B. "Sufficient Transformation" Standard with Meridian Sole Discretion
Meridian's proposal that Quorum be prohibited from commercializing Derived Data unless Meridian determines, in its sole discretion, that the data has been "sufficiently transformed" creates what is functionally a veto power over Quorum's ability to bring QuorumScore™ outputs to market as a B2B product. Under this provision, Meridian would have the unilateral authority to block any third-party licensing of QuorumScore™ analytics, regardless of the degree to which those outputs have been transformed, blended, or de-identified.
This concern is not hypothetical. QuorumScore™ outputs are, by design, the product of a complex multi-source blending process that integrates Meridian credit data with Quorum's own claims data, policy data, and telematics data. The resulting scores and models are several layers of transformation removed from any individual Meridian data element. Nevertheless, because the "sufficient transformation" standard is undefined and is subject to Meridian's sole discretion, Meridian could take the position that any output generated by a model that was trained on Meridian data — or that uses Meridian data as one of its inference-time inputs — has not been "sufficiently transformed," thereby blocking commercialization indefinitely.
The lack of objective criteria for what constitutes "sufficient transformation" compounds this risk. As we analyzed in the separate May 2023 memorandum, the concept of "transformation" in the context of machine-learning model outputs is technically complex and does not lend itself to a single subjective assessment. Reasonable experts could disagree about whether a given model output is "traceable to" a particular data input, particularly where the model has been trained on millions of records from multiple sources and produces outputs that reflect aggregate statistical patterns rather than identifiable individual data elements. By reserving sole discretion to itself, Meridian effectively eliminates any possibility of resolving such disagreements through objective analysis or expert determination.
Impact Assessment: The "sufficient transformation" standard, as currently drafted, creates commercial uncertainty that would chill investment in the QuorumScore™ B2B licensing pipeline and would give Meridian ongoing leverage to block or delay Quorum's commercialization strategy for competitive or strategic reasons unrelated to legitimate data-protection concerns.
C. 15% Revenue Share on Derived Data Commercialization
Meridian's demand for a 15% revenue share on net revenues attributable to commercialized Derived Data would impose a significant ongoing cost on Quorum's B2B licensing of QuorumScore™ outputs. On Quorum's projected $40–50 million in annual B2B revenue by 2027, this revenue share would cost Quorum approximately $6–7.5 million per year. This cost would be in addition to the $12.3 million annual license fee that Quorum is paying under the MDLA for access to Meridian's raw data products — bringing the total annual cost of the Meridian relationship, with respect to QuorumScore™ alone, to approximately $18.3–$19.8 million.
The revenue share is disproportionate to Meridian's contribution to the value of QuorumScore™ outputs. While Meridian's credit data is one input into the QuorumScore™ model, it is by no means the sole or even the predominant value driver. Quorum's first-party data assets — comprising over fifteen years of claims experience, detailed policy data, and telematics data — represent a data corpus that is not available from Meridian or any other third-party provider and that constitutes a significant portion of the input data used to train and operate QuorumScore™. Moreover, the proprietary algorithmic innovations, feature engineering processes, and model architecture that Quorum's data science team has developed represent independent intellectual capital that transforms raw data inputs (from all sources, not just Meridian) into differentiated analytical outputs. A 15% revenue share fails to account for this value distribution and effectively treats QuorumScore™ as if it were merely a repackaging of Meridian's data rather than an independent, proprietary analytics product.
Impact Assessment: The revenue share significantly reduces the economic returns on Quorum's data-science investment and B2B licensing strategy, undermining the business case for the QuorumScore™ licensing program and potentially rendering certain segments of the planned B2B business uneconomical.
D. Derivative Work Classification
Underlying Meridian's specific commercial demands is a legal theory that all Derived Data constitutes a "derivative work" under Section 101 of the Copyright Act and therefore remains subject to Meridian's underlying intellectual property rights in the raw data from which it was derived. This theory, if accepted — whether by contractual stipulation in the MDLA or through judicial determination in a subsequent dispute — would provide Meridian with an expansive basis for asserting control over virtually all QuorumScore™ outputs.
As Calderwood & Harkness analyzed in the January 2023 memorandum on machine-learning model outputs (copyright and trade secret analysis), the application of the "derivative work" framework to the outputs of machine-learning models is legally aggressive and analytically unsupported in the context of multi-source, algorithmically transformed analytics like QuorumScore™. The key analytical distinction is between:
(i) Model weights and parameters, which are the internal numerical values learned by the model during the training process. While these values are statistically influenced by the training data (including Meridian's credit data), they do not "reproduce" or "transform" any individual training data element in the sense contemplated by the Copyright Act's derivative work definition. Model weights are the distilled statistical patterns extracted from training data, not copies or transformations of that data.
(ii) Live predictions and inference-time outputs, which are generated when the trained model is applied to new, real-time data inputs — including applicant data, Quorum's first-party data, and other non-Meridian sources. These outputs are independently generated by the model's inference engine and do not incorporate, reproduce, or display Meridian's raw credit data in any recognizable form. A QuorumScore™ risk score generated for an individual underwriting decision is a product of the model's learned patterns applied to new data, not a reproduction or transformation of any individual Meridian data record.
This distinction is critical. QuorumScore™ live predictions for individual underwriting decisions use real-time applicant data, Quorum's first-party claims and policy data, and the trained model — not Meridian's raw credit data in any identifiable or reconstructible form. To classify these outputs as "derivative works" of Meridian's data would stretch the Copyright Act's derivative work framework well beyond its intended scope and would effectively grant data licensors perpetual control over all downstream analytical outputs generated by any licensee that uses the licensed data as one input among many in a complex analytical process.
Impact Assessment: If Meridian's derivative-work theory is accepted contractually through the MDLA — regardless of whether it would withstand judicial scrutiny — it would extend Meridian's intellectual property claims to virtually all QuorumScore™ outputs, regardless of how attenuated the connection between those outputs and Meridian's raw data. This would create a contractual basis for Meridian to assert the license-back, revenue share, and transformation restrictions discussed above, even with respect to outputs that bear no meaningful resemblance to Meridian's underlying data.
IV. Trade Secret Protection Strategy
A. Current Trade Secret Safeguards
QuorumScore™'s model weights, feature engineering processes, training data configurations, and algorithm parameters are currently maintained as trade secrets under Quorum's enterprise information security program. This classification is appropriate, as these elements derive independent economic value from the fact that they are not generally known to or readily ascertainable by Quorum's competitors through proper means, and because Quorum has taken reasonable efforts to maintain their secrecy — the two essential requirements for trade secret protection under both the Defend Trade Secrets Act ("DTSA"), 18 U.S.C. § 1836 et seq., and applicable state trade secret law.
The specific safeguards that Quorum has implemented to protect the secrecy of QuorumScore™'s trade-secret-protected elements include the following:
Access Controls. Access to QuorumScore™ model weights, training configurations, feature engineering code, and algorithm parameters is governed by role-based access controls within Quorum's information technology environment. Access is limited to members of Quorum's data science and underwriting analytics teams who have a demonstrated business need for access. All access is logged, and access logs are subject to periodic audit by Quorum's internal audit function. Multi-factor authentication is required for access to the production systems and code repositories in which QuorumScore™'s trade-secret-protected elements reside.
Confidentiality Obligations. All Quorum employees with access to QuorumScore™ trade secrets have executed confidentiality obligations, either as part of their employment agreements or through standalone non-disclosure agreements. These obligations require employees to maintain the confidentiality of proprietary information, prohibit unauthorized disclosure, and survive the termination of employment. Contractors and consultants who are granted access to QuorumScore™ systems or information are required to execute non-disclosure agreements prior to receiving access.
Information Classification. QuorumScore™'s trade-secret-protected elements are classified at the highest tier of Quorum's information classification scheme ("Restricted — Trade Secret"), which imposes the most stringent handling, storage, and transmission requirements under Quorum's information security policies. Documents and data sets containing trade-secret-protected information are marked accordingly, and personnel handling such materials are required to follow enhanced security protocols.
SOC 2 Type II Certification. Quorum's systems environment — including the infrastructure in which QuorumScore™ operates — is the subject of a SOC 2 Type II report issued by Felton Thorpe & Co. LLP for the period ending December 31, 2023. This independent audit provides third-party assurance that Quorum's security controls meet the Trust Services Criteria established by the American Institute of Certified Public Accountants for security, availability, processing integrity, confidentiality, and privacy. The SOC 2 Type II report is relevant to the trade secret analysis because it evidences Quorum's ongoing commitment to maintaining reasonable security measures — a factor that courts consider in assessing whether a party has taken "reasonable efforts" to maintain secrecy under both the DTSA and state trade secret statutes.
B. Trade Secret Risks Under Meridian's MDLA Positions
The trade secret protection that QuorumScore™ currently enjoys is directly threatened by several of Meridian's MDLA positions. Most critically, a perpetual license-back to Meridian on Derived Data that encompasses QuorumScore™ outputs would constitute a voluntary disclosure of trade-secret-protected information to a third party, with potentially irreversible consequences for the viability of QuorumScore™'s trade secret status.
Under both the DTSA and the Uniform Trade Secrets Act (as adopted in Connecticut, where Quorum is headquartered), trade secret protection requires that the information derive independent economic value from not being generally known or readily ascertainable. Voluntary disclosure of QuorumScore™ outputs to Meridian — a sophisticated data analytics company with the technical capability to analyze those outputs and reverse-engineer the underlying model — would undermine Quorum's ability to argue that the information is not "readily ascertainable" by competitors. Even if the license-back is nominally subject to confidentiality obligations imposed on Meridian, the practical reality is that providing Meridian with ongoing access to QuorumScore™ outputs over a period of years would enable Meridian's data scientists to develop a detailed understanding of QuorumScore™'s behavior, sensitivities, and outputs. Through techniques such as model inversion, membership inference, and systematic probing with varied inputs, a technically sophisticated recipient of model outputs can reconstruct significant aspects of the underlying model — including its feature importance rankings, decision boundaries, and approximate parameter values.
The risk is compounded by the fact that Meridian operates a competing analytics business and has both the incentive and the technical capacity to exploit any insights gained from QuorumScore™ outputs. Even absent deliberate reverse engineering, the mere availability of QuorumScore™ outputs within Meridian's organization creates a risk of inadvertent disclosure, employee misuse, or organizational leakage — risks that Quorum cannot effectively monitor or control within Meridian's corporate environment.
Risk Assessment: Loss of trade secret protection for QuorumScore™ would be effectively irreversible. Unlike patent or copyright protection, which exists as a matter of law regardless of the information's secrecy, trade secret protection depends on the information remaining secret. Once trade-secret-protected information enters the public domain or is shared without adequate protections, trade secret status cannot be restored. The loss of trade secret protection would remove an entire layer of Quorum's IP protection strategy for QuorumScore™ and would leave Quorum reliant on patent protection (which is pending and not yet granted) and contractual restrictions (which are enforceable only against parties who have agreed to them) as the primary means of protecting the model's proprietary value.
C. Recommended Safeguards
To preserve QuorumScore™'s trade secret protections in the context of the MDLA relationship, Calderwood & Harkness recommends the following safeguards:
MDLA Confidentiality Provisions. The MDLA must contain express provisions establishing that QuorumScore™ and its outputs are Quorum's "Confidential Information" under the agreement, distinct from and not subject to the Derived Data provisions of Article VII. This classification should be reinforced with specific contractual language providing that Meridian receives no right to access, use, analyze, reverse-engineer, or derive insights from QuorumScore™ outputs under any circumstance.
Limitation on Data Sharing. Any data shared with Meridian in the course of the licensing relationship — whether for troubleshooting, data-quality validation, integration testing, or other operational purposes — must be limited to non-trade-secret inputs and must not include QuorumScore™ model outputs, model weights, algorithmic parameters, feature engineering logic, or any other trade-secret-protected elements. Protocols should be established to ensure that Quorum's data engineering team does not inadvertently transmit QuorumScore™ outputs to Meridian in the course of routine data exchanges. Priya Nair's team should implement technical safeguards, such as data-flow monitoring and outbound filtering, to enforce this restriction at the system level.
Data Security Addendum. The Data Security Addendum ("DSA") that accompanies the MDLA should include provisions requiring Meridian to implement access controls, encryption at rest and in transit, use restrictions, and audit rights with respect to any Quorum Confidential Information to which Meridian obtains access in the course of the relationship. While the primary objective is to prevent Meridian from accessing QuorumScore™ information at all, the DSA should provide a backstop of technical and organizational safeguards in the event that any such information is inadvertently disclosed.
Regulatory Cross-Reference. The confidentiality and data-security provisions of the MDLA should be coordinated with the regulatory requirements applicable to Quorum's data-handling practices. Claudia Reeves, VP of Regulatory & Compliance at Quorum, should be consulted regarding state insurance data-security regulations — including the requirements of the Connecticut Insurance Data Security Law (Conn. Gen. Stat. § 38a-38) and the New York Department of Financial Services Cybersecurity Regulation (23 NYCRR 500) — that impose obligations on Quorum with respect to the protection of confidential and proprietary information. These regulatory requirements provide additional support for restricting Meridian's access to QuorumScore™ trade secrets and can serve as a framework for the specific security controls required under the DSA.
V. Patent Protection Strategy
A. Status of USPTO Application
As noted in Section II.B above, QuorumScore™ is the subject of a pending patent application before the United States Patent and Trademark Office, Serial No. 97/482,116. The application was prepared and filed by Caldwell & Strauss LLP under the direction of Angela Firth and covers three principal aspects of the QuorumScore™ system: (1) the model architecture, including the specific neural network topology and ensemble methodology; (2) the scoring methodology, including the proprietary approach to multi-source data blending and risk quantification; and (3) certain algorithmic innovations developed by Quorum's data science team, including novel feature engineering techniques and optimization methods.
The application is currently in the pending stage and has not yet been examined by the USPTO. No office actions have been received, and no examiner has been assigned. The timeline for examination depends on the current backlog in the relevant technology class, which — for applications related to artificial intelligence and machine-learning systems — has historically ranged from eighteen to thirty months from the filing date. The application was filed in 2022, which means that examination could begin as early as late 2023 or as late as mid-2025 under normal processing timelines.
B. Strategic Considerations in Context of MDLA
The pendency of the patent application has significant implications for Quorum's negotiating position in the MDLA negotiation. A granted patent would provide Quorum with an independent, legally enforceable basis for asserting that QuorumScore™ and its outputs are Quorum's proprietary technology, separate from and not subject to any copyright claims that Meridian may assert over its raw data. This is because patent rights are independent of copyright: a patented method or system is protected regardless of the data inputs used in its operation, and the outputs of a patented system are the proprietary outputs of the patent holder, not derivative works of the data that the system processes.
Patent protection is particularly important in the context of the planned B2B licensing business. Third-party licensees of QuorumScore™ analytics will require assurance that Quorum has clear intellectual property ownership and freedom to operate — that is, that no third party has the ability to challenge Quorum's right to license QuorumScore™ outputs or to demand a revenue share on revenues generated by licensees' use of those outputs. A granted patent would provide strong support for these assurances and would significantly reduce the legal and commercial risk perceived by prospective B2B licensees.
Conversely, if the MDLA grants Meridian a license-back or revenue share on Derived Data that encompasses QuorumScore™ outputs, the commercial value of the patent (and the planned B2B licensing business) would be significantly diminished. Prospective licensees would face the prospect of Quorum's rights being encumbered by Meridian's claims, potentially subjecting them to Meridian's approval requirements or revenue-sharing demands as a condition of their license from Quorum. This would create an unacceptable level of commercial and legal uncertainty that would likely deter sophisticated licensees from entering into agreements with Quorum.
Recommendation: Calderwood & Harkness recommends that Quorum coordinate with Caldwell & Strauss LLP to evaluate the feasibility of filing a petition for Track One prioritized examination with the USPTO. Track One prioritized examination, established under 37 C.F.R. § 1.102(e), provides for a final disposition within twelve months of the filing of the petition in exchange for an additional fee (currently $4,000 for large entities). If Track One examination is available — and if the application meets the eligibility requirements — expedited prosecution could result in a granted patent prior to the execution of the MDLA, materially strengthening Quorum's negotiating position by demonstrating issued patent rights in the QuorumScore™ system. Even if Track One is not available or is not granted, any acceleration of the examination timeline would be strategically beneficial.
C. Patent-Related Contractual Provisions
Regardless of the status of the patent application at the time the MDLA is executed, Calderwood & Harkness recommends that the MDLA contain the following patent-related contractual provisions:
Quorum IP Ownership Representation. The MDLA should contain a clear representation by Quorum that it owns and retains all patent rights — including rights under both issued patents and pending patent applications — in QuorumScore™, its model architecture, scoring methodology, algorithmic innovations, and all outputs generated by the QuorumScore™ system. This representation should be drafted broadly enough to encompass both the currently pending application (Serial No. 97/482,116) and any continuation, divisional, or related applications that Quorum may file in the future.
Meridian Acknowledgment. Meridian should acknowledge in the MDLA that no rights in Quorum's patented or patent-pending technology are granted, assigned, or licensed to Meridian under the MDLA. This acknowledgment should be explicit and unambiguous, foreclosing any argument that Meridian's receipt of Derived Data under the MDLA constitutes an implied license to QuorumScore™ or its underlying technology.
Patent Exclusion from Derived Data. A specific exclusion should state that QuorumScore™ outputs, to the extent covered by Quorum's patent claims (whether issued or pending), are not "Derived Data" as defined in the MDLA and are not subject to any of the provisions of Article VII (Derived Data) of the agreement. This provision would operate as a belt-and-suspenders complement to the broader QuorumScore™ carve-out recommended in Section VI.A below, providing an independent basis for excluding QuorumScore™ from Meridian's Derived Data claims.
VI. Recommended Contractual Provisions in the MDLA
Based on the foregoing analysis, Calderwood & Harkness recommends that the following contractual provisions be incorporated into the MDLA. These recommendations are presented in order of priority and should be treated as essential elements of any final agreement. We are prepared to draft the specific contractual language for inclusion in the next MDLA redline upon instruction from Quorum's Office of General Counsel.
A. Definition of "Derived Data" — QuorumScore™ Carve-Out
The most critical contractual safeguard is a clear and express carve-out of QuorumScore™ and its outputs from the MDLA's definition of "Derived Data." Under the current draft MDLA, "Derived Data" is defined to encompass any data, analytics, models, scores, indices, or other outputs that "incorporate, are based on, or are traceable to" Meridian's raw data. This definition is sufficiently broad that it could be construed to encompass QuorumScore™ outputs, on the theory that QuorumScore™ is "based on" or "traceable to" the Meridian credit data used as one input in the model's training process.
Calderwood & Harkness recommends that the MDLA's definition of "Derived Data" be amended to expressly exclude any outputs of Quorum's proprietary models, including QuorumScore™, where the following conditions are met:
(i) the model was trained using data from multiple sources, of which Meridian's data is only one;
(ii) the output does not reproduce, display, or contain any Meridian raw data in identifiable or individually attributable form; and
(iii) the output is generated using non-Meridian inputs at inference time — that is, the model produces its output by processing real-time applicant data, Quorum first-party data, or other non-Meridian data through the trained model, rather than by extracting, repackaging, or transforming Meridian raw data.
As an alternative or complementary approach, Calderwood & Harkness recommends that the MDLA define a category of "Quorum Proprietary Outputs" or "Excluded Outputs" that expressly encompasses QuorumScore™ predictions, risk scores, propensity models, enriched datasets, and any other analytics outputs generated by Quorum's proprietary models. The MDLA should then state, in unambiguous terms, that Quorum Proprietary Outputs are not subject to Article VII (Derived Data), are not subject to any license-back, revenue share, or transformation requirement, and are Quorum's exclusive intellectual property without limitation.
The specific drafting approach should be determined in consultation with Quorum's in-house team, but the substantive result must be the same: QuorumScore™ outputs must be excluded from the Derived Data framework, fully and unconditionally.
B. No License-Back on QuorumScore™
The MDLA must contain an express provision stating that Meridian receives no license, right, title, or interest — perpetual, term-limited, royalty-free, or otherwise — in QuorumScore™, its model architecture, model weights, training data configurations, or any outputs generated by QuorumScore™. This restriction must apply regardless of whether QuorumScore™ was trained using Meridian data as one of its inputs.
The rationale for this provision is straightforward. QuorumScore™ reflects Quorum's independent and substantial investment in data science talent, first-party data assets (including claims data, policy data, and telematics data), and proprietary algorithmic innovation. The fact that Meridian's credit data is one of several inputs used to train the model does not entitle Meridian to a license over the model's outputs any more than the manufacturer of a camera is entitled to a license over the photographs taken with that camera. Meridian's contribution to the QuorumScore™ ecosystem is the provision of credit data, for which Quorum is compensating Meridian at a rate of $60 million over five years. That compensation fully exhausts Meridian's entitlement to value from the data it provides.
C. No Revenue Share on QuorumScore™ Commercialization
The MDLA should expressly exclude QuorumScore™ and its outputs from any revenue-sharing arrangement with Meridian. This exclusion is critical to preserving the economics of the projected $40–50 million B2B licensing revenue line by 2027. As analyzed in Section III.C above, a 15% revenue share on QuorumScore™ B2B revenues would cost Quorum approximately $6–7.5 million per year and would significantly reduce the return on Quorum's data-science investment.
We note that Quorum continues to resist any revenue-sharing arrangement on Derived Data generally, and we support that position. However, if Quorum ultimately agrees to some form of revenue share on other categories of Derived Data — for example, on analytics outputs that are more closely traceable to Meridian's raw data and less the product of Quorum's independent intellectual investment — it is imperative that QuorumScore™ be carved out of any such obligation. The carve-out should be explicit, specific, and unambiguous, and it should be cross-referenced to the definition of "Quorum Proprietary Outputs" or "Excluded Outputs" described in Section VI.A above.
D. Objective Transformation Standard
If the MDLA includes any "sufficient transformation" requirement for third-party commercialization of Derived Data — whether applied to QuorumScore™ outputs or to other categories of Derived Data — the standard must be objective and jointly determined, not subject to Meridian's sole discretion. Meridian's current proposal, which reserves sole discretion to Meridian, is commercially unworkable and gives Meridian an effective veto over Quorum's business strategy. No sophisticated enterprise customer would accept such a provision, and Quorum should not be the first to do so.
Calderwood & Harkness recommends that the MDLA adopt the criteria set forth in the June 2023 Derived Data Commercialization Protocol draft that was proposed by Quorum at the third negotiation session. These criteria include:
• Statistical irreversibility tests: quantitative assessments, performed using accepted statistical methodologies, to confirm that the Derived Data output cannot be used to reconstruct or reverse-engineer individual Meridian data records;
• K-anonymity thresholds: minimum anonymization standards ensuring that any data element in the Derived Data output relates to a group of no fewer than a specified number of individuals (with the threshold to be agreed by the parties);
• Independent expert certification: a process by which a qualified, independent third-party expert — selected by mutual agreement of the parties or from a pre-approved roster — certifies that the Derived Data output meets the agreed transformation standard; and
• Other measurable benchmarks: such additional quantitative or qualitative criteria as the parties may agree, provided that they are defined with sufficient specificity to permit objective evaluation.
With respect to QuorumScore™ outputs specifically, Calderwood & Harkness recommends that the MDLA provide that QuorumScore™ outputs are deemed to satisfy any applicable transformation standard as a matter of contract. This deemed-satisfaction provision is justified by the inherent nature of QuorumScore™: the model's multi-source blending process, algorithmic transformation, and inference-time use of non-Meridian inputs ensure that QuorumScore™ outputs are several layers of transformation removed from any individual Meridian data element and cannot be used to reconstruct Meridian's raw data.
E. Post-Termination Treatment of QuorumScore™
The MDLA should expressly provide that upon termination or expiration of the agreement — whether by lapse, non-renewal, or termination for cause — Quorum retains all rights to QuorumScore™, including the trained model (with all model weights and parameters), all outputs generated during the license term, and all derivative models and improvements developed by Quorum during the term.
Under the current draft MDLA, Quorum is required to return or destroy all Meridian raw data within ninety (90) days of termination. This obligation is reasonable with respect to Meridian's raw data itself, and Quorum does not object to it. However, the return-or-destroy obligation must not extend to QuorumScore™'s model weights, parameters, or outputs, even if those weights were influenced by training on Meridian data during the license term. The model weights are not copies or reproductions of Meridian's data; they are the distilled statistical patterns learned by the model from all of its training data sources, including Quorum's own first-party data. Requiring Quorum to delete a trained model because one of its training data sources was Meridian's credit data would be commercially untenable — it would require Quorum to discard years of data-science investment and start from scratch — and is legally unsupported, as discussed in Section III.D above.
This is a disputed point. Meridian's current position on post-termination treatment of Derived Data is formally unresolved, and there are indications that Meridian may seek to require the deletion of any model trained on its data. Calderwood & Harkness recommends that Quorum take a firm position on this issue and treat it as a non-negotiable element of the MDLA. The post-termination retention of QuorumScore™ is essential to preserving the continuity and value of the B2B licensing business, which cannot be interrupted or unwound in the event of a Meridian contract non-renewal.
F. Representations and Warranties
Calderwood & Harkness recommends that the MDLA include the following representations and warranties to reinforce Quorum's IP ownership position:
By Quorum: Quorum should represent and warrant that it is the sole owner of all intellectual property rights in QuorumScore™, including all patent, copyright, trade secret, and trademark rights; that it has the right to use, sublicense, and commercialize QuorumScore™ outputs without the consent of any third party; and that the execution and performance of the MDLA does not and will not conflict with or impair Quorum's ownership or exploitation rights in QuorumScore™.
By Meridian: Meridian should represent and warrant that it makes no claim of ownership, license, or other interest in QuorumScore™ or any outputs thereof; that it will not assert any intellectual property claim against Quorum or any third-party licensee of Quorum based on QuorumScore™ outputs; and that the Derived Data provisions of the MDLA do not apply to QuorumScore™ or any outputs thereof, as set forth in the carve-out provisions described in Section VI.A above.
Survival: Both sets of representations and warranties should expressly survive the termination or expiration of the MDLA, ensuring that Meridian's acknowledgment of Quorum's IP ownership in QuorumScore™ remains effective regardless of the continued existence of the contractual relationship.
VII. Interaction with Standish National Bank Covenant
An additional consideration that bears on the MDLA negotiation — and specifically on the QuorumScore™ IP protection strategy — is the negative covenant contained in Quorum's existing credit agreement with Standish National Bank, N.A.
Quorum maintains a $1.2 billion revolving credit facility with Standish National Bank, N.A. ("Standish"), under which Philip Yardley serves as Relationship Manager. Section 7.12(b) of the Amended and Restated Credit Agreement, dated June 15, 2021, between Quorum and Standish (the "Credit Agreement"), contains a negative covenant requiring Quorum to obtain the prior written consent of the lender before entering into any material intellectual property license with an aggregate value exceeding $25,000,000. The MDLA, with an aggregate value of $60,000,000 over the initial five-year term, exceeds this threshold by $35,000,000 — representing 240% of the consent threshold. Lender consent is therefore required before Quorum can execute the MDLA.
The lender consent process is being managed by Marcus Weill from Quorum's treasury function, in coordination with Quorum's Office of General Counsel. Calderwood & Harkness recommends that the materials submitted to Standish in connection with the consent request include a clear description of Quorum's retained intellectual property rights in QuorumScore™. Specifically, the consent materials should explain that the MDLA is a data-licensing arrangement under which Quorum receives access to Meridian's credit data products in exchange for license fees, and that the MDLA does not grant Meridian any rights in QuorumScore™ or any other Quorum proprietary technology. This description will assure Standish that the MDLA does not encumber or transfer Quorum's core proprietary technology — an assurance that Standish will likely require before granting consent, given the significance of QuorumScore™ to Quorum's enterprise value and growth strategy.
The corollary is also important: if the MDLA were to grant Meridian a license-back or revenue share on QuorumScore™ outputs, Standish could reasonably view this as a further encumbrance on Quorum's intellectual property portfolio. A perpetual, royalty-free license-back on QuorumScore™ outputs to a third party — particularly a third party that operates a competing analytics business — could be characterized as a material impairment of the intellectual property assets that serve as part of the collateral package supporting the credit facility. This characterization could complicate the consent process, potentially leading to additional conditions, covenants, or consent fees imposed by Standish, or — in a worst case — to a refusal of consent that would prevent Quorum from executing the MDLA on the proposed terms.
Ensuring that the MDLA preserves Quorum's unencumbered IP rights in QuorumScore™ is therefore not only a matter of commercial strategy but also a practical prerequisite for obtaining the lender co