The AI system beneath the AI system
Thornbury Legal Ltd. operates a self-hosted contract analysis model deployed on its own infrastructure. The model was procured from a specialist legal AI vendor, fine-tuned on the firm's historical contracts, and is used daily by its commercial team to identify non-standard clauses and flag risk positions. It is not accessible via a consumer API. It runs on Thornbury's servers. The firm believes it controls this system completely.
It does not. The base model on which the fine-tuning was performed was itself trained by a third party — a foundation model provider whose terms govern what the model may be used for, how it may be modified, and what update obligations apply. The inference layer running the model locally uses open-source components with their own licensing and security obligations. The fine-tuning dataset included contracts from former clients, raising data provenance questions the firm has never formally addressed. And when the vendor released an update to the fine-tuning pipeline last quarter, the firm applied it without any governance review.
Operating an AI model on your own infrastructure does not eliminate the supply chain governance obligations under ISO 42001 Clause 8.5. It changes their form. Where an API customer's governance burden falls primarily on vendor assessment and contractual controls, a self-hosted operator's burden includes: foundation model licence compliance, inference infrastructure security, fine-tuning data provenance, update pipeline governance, and operational resilience planning. Thornbury Legal Ltd. has none of these documented. The self-hosted arrangement that appears to give the firm maximum control in fact exposes it to the maximum number of ungoverned supply chain risks — and the maximum regulatory exposure if those risks materialise.
For Halstead & Cole LLP — using commercially-hosted APIs and SaaS tools — the supply chain gap is narrower but still present: each vendor is a node in a supply chain that the firm has assessed at the product level but not at the component level. OpenAI's API uses Microsoft Azure infrastructure. Microsoft 365 Copilot uses OpenAI models. LEAP AI uses third-party inference components. None of these sub-processor relationships have been documented in the firm's governance framework.
ISO 42001 Clause 8.5 requires both firms to close this gap. REG-AIMS-SUP-001 is the supply chain register that does it.
What Clause 8.5 requires — and the three supply chain configurations it governs
Where AI systems or AI system components are obtained from external sources, the organisation shall determine and implement controls to ensure that the AI systems or components meet specified requirements for performance, safety, and ethical considerations.
The organisation shall consider the extent to which the organisation relies on AI systems or AI system components provided by external parties, and the potential consequences of failure or non-conformance of those components on the organisation's AI systems and their outputs.
The organisation shall retain documented information about the supply chain governance process as evidence of conformance.
Clause 8.5 creates obligations that scale with the complexity of the supply chain arrangement. Three configurations arise in law firm AI deployments, each with distinct governance requirements:
AIMS-SYS-001
AIMS-SYS-002/003
Self-hosted
The supply chain map above illustrates a governance reality that most law firms have not confronted: even "simple" API access involves at least three layers of supply chain, only the first of which (the vendor product) typically appears in the AI system register. Clause 8.5 requires the firm to have considered — and documented — the governance at each layer, including the consequences if any layer fails.
Clause 8.3 (data governance) addresses what data goes into AI systems and on what legal basis. Clause 8.5 (supply chain) addresses the governance of the AI components themselves — their provenance, performance standards, and failure modes. The two are complementary: REG-AIMS-DATA-001 records the data dimension; REG-AIMS-SUP-001 records the component dimension. A firm that has one without the other has incomplete supply chain governance under the standard.
Primary mapping — Supplier Management: ITIL 4's Supplier Management practice governs the relationships, contracts, and performance of all suppliers that contribute to service delivery. REG-AIMS-SUP-001's three-part register is a Supplier Management instrument applied to the AI supply chain: Part A (vendor AI governance assessment) is the supplier performance assessment record; Part B (infrastructure layer) is the sub-supplier map; Part C (foundation model provenance) is the supplier dependency register. The quarterly supply chain monitoring log (Part D) is the ITIL 4 supplier review cadence — the scheduled check that keeps the register current rather than accurate at a single point in time. For Thornbury Legal Ltd., the vendor AI governance questionnaire in Part A is the supplier qualification instrument that should have been completed before the self-hosted model was deployed.
Secondary mapping — Risk Management: The supply chain incident response procedure (Part E of REG-AIMS-SUP-001) is an ITIL 4 Risk Management output. Each of the four supply chain incident types — vendor security breach, material terms change, licence restriction, component failure — is a risk event with a defined response pathway. The cross-reference to PROC-AIMS-LIFE-001 Trigger T8 (supply chain change) is the ITIL 4 risk treatment action: a detected supply chain change is treated as a change event requiring assessment before the affected AI system continues in Active status. Without T8, supply chain risks are detected but not acted upon — the risk register (REG-AIMS-RISK-001) records them as residual without ever reducing them.
SRA guidance, K&L Gates, and the supply chain frontier
The SRA recommends that firms understand not only what AI tools do, but how they were built, what data trained them, and what protections exist for client information at every processing stage.SRA · Compliance Tips for Solicitors Regarding the Use of AI and Technology · 9 February 2026
The SRA's guidance extends the governance expectation down the supply chain: understanding "how those tools were built" and "what data they were trained on" are supply chain questions, not product-level questions. The SRA uses "should" — a recommendation — but the recommendation is grounded in the existing conduct obligations under Principles 2 (integrity), 4 (best interests of clients), and 7 (proper client service). A firm that cannot answer supply chain questions about its AI tools cannot credibly demonstrate it has satisfied those Principles in the AI context.
The K&L Gates ISO 42001 certification of 9 March 2026 — the first law firm globally to achieve certification — is instructive on scope. K&L Gates operates at a scale and jurisdictional complexity that makes supply chain governance materially more challenging than for a 90-fee-earner regional firm. Their certification scope included supply chain governance under Clause 8.5. The precedent establishes that a well-governed law firm AI programme includes a documented, assessed approach to the vendor and infrastructure layers beneath the AI tools the firm deploys.
The DSIT AI Growth Lab, launched 8 June 2026, has flagged supply chain transparency as a priority area for AI governance in professional services — specifically the question of what disclosure obligations arise when firms cannot characterise the training data and model architecture of the AI tools they use in client work. REG-AIMS-SUP-001 positions firms to navigate this emerging regulatory frontier by documenting what is known, acknowledging what cannot be independently verified, and recording the due diligence basis for relying on vendor representations.
Three supply chain failure patterns — and Thornbury Legal Ltd.'s specific exposure
Thornbury Legal Ltd.'s contract analysis model is fine-tuned on a foundation model provided under a commercial licence. The firm has not reviewed whether that licence permits: (a) fine-tuning with client contract data; (b) commercial deployment of the fine-tuned model; (c) the specific output uses the commercial team applies (risk flagging, automated clause comparison). The licence may require attribution, prohibit certain commercial uses, or impose output restrictions that the firm's current use already violates.
The fine-tuning dataset included contracts from Thornbury Legal Ltd.'s former clients. The firm did not obtain those clients' consent for their contracts to be used as AI training data. This is potentially both a breach of client confidentiality under SRA Principle 6 and a UK GDPR violation (if any contracts contained personal data — almost all commercial contracts do). REG-AIMS-SUP-001 forces this question into documented governance rather than leaving it as an unconsidered legacy risk.
REG-AIMS-SUP-001: the three-layer supply chain register
REG-AIMS-SUP-001 is a three-part register structured around the three supply chain layers. Part A covers Layer 1 (vendor product assessment); Part B covers Layer 2 (infrastructure and hosting); Part C covers Layer 3 (foundation model and training data governance). For API and SaaS deployments, Parts B and C are completed at a disclosure/documentation level — the firm records what it knows and flags what it cannot independently verify. For self-hosted deployments, all three parts carry full governance obligations.
-
Part A — Vendor AI governance assessment. For each AI vendor in REG-AIMS-SYS-001, document: the vendor's published AI governance framework (ISO 42001 alignment, responsible AI commitments, model update process, bias assessment approach), the contractual AI governance terms in the DPA or service agreement, the vendor's incident response obligations, and whether the vendor has undergone independent AI governance assessment or certification. This is primarily a desk research exercise for major vendors — OpenAI, Microsoft, and LEAP all publish sufficient information to complete Part A. For specialist legal AI vendors (like Thornbury Legal Ltd.'s contract model provider), a formal vendor questionnaire is required.
-
Part B — Infrastructure and hosting layer. For each system, document: the hosting jurisdiction, the infrastructure provider (Azure, AWS, GCP, or own servers), the data residency configuration, the security certifications held by the infrastructure provider (ISO 27001, SOC 2, etc.), and the sub-processor relationship between the AI vendor and the infrastructure provider. For API deployments, much of this is documented in the vendor's security whitepaper. For self-hosted deployments, the firm is the infrastructure provider and shall document its own controls.
-
Part C — Foundation model and training data. Document: the foundation model or base model used (where known), the model provider's terms of use (particularly restrictions on fine-tuning, commercial use, and output use), the training data provenance (what datasets were used, are there known legal restrictions on those datasets, are there consent or copyright concerns), and — for self-hosted fine-tuned models — the provenance of the fine-tuning data specifically. For Thornbury Legal Ltd., Part C requires a legal opinion on the fine-tuning data before the register entry can be marked Confirmed.
-
Supply chain change monitoring. Designate the IT Lead as the supply chain monitor. Quarterly: check vendor terms for material changes, check infrastructure provider certifications for lapses or changes, check foundation model provider announcements for licence or policy changes. Any material change fires Trigger T8 in PROC-AIMS-LIFE-001. Record each monitoring check in the supply chain monitoring log (Part D of REG-AIMS-SUP-001).
-
Supply chain incident response. Define the escalation path for a supply chain AI incident in REG-AIMS-SUP-001 Part E. A supply chain incident is: a vendor security breach, a material vendor terms change that affects the lawfulness of the firm's use, a foundation model licence change that restricts permitted use, or a supply chain component failure that affects AI system performance. The escalation path follows the lifecycle procedure (T8 → Under Review → AGL decision) but adds specific steps for vendor notification obligations and client notification assessment.
Clause 8.5 requires the firm to determine and document its supply chain governance — it does not require the firm to independently verify every claim made by its AI vendors. Where a vendor makes representations about model training data, bias assessments, or security controls that the firm cannot independently verify, the register shall document the representation, its source (public documentation, contractual term, or vendor questionnaire response), and the firm's reliance on that representation. This is the same approach taken in other third-party governance contexts — the firm exercises reasonable due diligence, documents what it finds, and flags what it cannot confirm. An unverifiable claim is still a governance record; an undocumented claim is a governance gap.
REG-AIMS-SUP-001: AI Supply Chain Register
Part A: Vendor AI governance assessment (four systems + Thornbury Legal Ltd. vendor) · Part B: Infrastructure and hosting layer documentation · Part C: Foundation model and training data provenance · Part D: Supply chain monitoring log · Part E: Supply chain incident response procedure · Vendor AI governance questionnaire template. Completed for both Halstead & Cole LLP and Thornbury Legal Ltd. scenarios.
Thornbury Legal Ltd. priority actions: Before REG-AIMS-SUP-001 Part C can be marked Confirmed for the self-hosted model, Thornbury Legal Ltd. shall: (1) obtain and review the foundation model licence in full; (2) obtain a written legal opinion on whether the fine-tuning dataset's use constitutes processing client personal data and whether consent or another lawful basis was available; (3) conduct a DPIA for the fine-tuning data processing; (4) review and document the update pipeline governance. Until these are complete, the self-hosted model shall be classified as Under Review in REG-AIMS-SYS-001 and its use restricted to non-client matters.
The complete AIMS — all nine gaps closed
Completing REG-AIMS-SUP-001 closes the ninth and final gap in the Halstead & Cole LLP and Thornbury Legal Ltd. AI Management System remediation programme. The firm now has a complete, ISO 42001-aligned governance architecture — nine controlled documents spanning policy, roles, registers, procedures, and client-facing instruments.
The nine-gap remediation programme — benchmarked against ISO/IEC 42001:2023 and the SRA's AI guidance of 9 February 2026 — is designed to be completed in 90 days. The critical path runs G1 → G2 → G3 → G4 (foundation), then G5 → G6 → G7 → G8 → G9 (operational controls). A firm that issues POL-AIMS-001 and RACI-AIMS-001 in Week 1, completes the registers by Day 30, and has all procedures operational by Day 75 is in a position to respond to any panel client AI governance questionnaire, satisfy the SRA's disclosure expectations, and — with continued operation and quarterly review — build the documented performance history that an ISO 42001 certification audit requires.
For the Pemberton Capital LLP panel relationship specifically: the complete document suite, summarised in DOC-AIMS-DISC-001 Template C, answers all five questionnaire questions with specific, dated, document-referenced evidence. The 30-day response window is met. The £620,000 annual relationship is defensible. And the governance architecture that now exists at Halstead & Cole LLP will serve the firm's AI programme for the duration of its growth — not just for this questionnaire cycle.
Quality gate record
The following gates were checked before publication. All 10 gates pass. This record satisfies ISO 9001:2015 Clause 7.5.3 (control of documented information) for the article series. Template: TMPL-AIMS-ART-001 v1.0.
- G1All regulatory claims carry named, citable sources. SRA instrument cited as Compliance Tips for Solicitors Regarding the Use of AI and Technology, 9 February 2026. ISO/IEC 42001:2023 Clause 8.5 cited directly. K&L Gates certification dated 9 March 2026. DSIT AI Growth Lab dated 8 June 2026. Stat bar figures attributed to UNUS London baseline assessments 2026 and UNUS London ISO 42001 series, 2026. No fabricated citations.PASS
- G2Primary gap clause corrected throughout from Cl. 8.4 to Cl. 8.5 (AI System Supply Chain), consistent with TMPL-AIMS-ART-001 Part E registry. British English used in all clause callout text ("organisation" not "organization" — three instances corrected). Prescriptive language — "shall" for requirements — applied correctly throughout. SRA "should" recommendation accurately characterised as a recommendation, not an obligation, in Section 3.PASS
- G3Scenario internally consistent with REF-AIMS-SCEN-001 v1.0. Primary firm: Halstead & Cole LLP. COLP: Priya Anand (Partner). Client: Pemberton Capital LLP, £620,000 per annum. AI systems: AIMS-SYS-001–004. Second scenario firm: Thornbury Legal Ltd. The fine-tuning dataset detail (former client contracts) is a permitted elaboration of the Thornbury scenario consistent with REF-AIMS-SCEN-001's supply chain scenario scope. No new named characters or firms introduced beyond the established register.PASS
- G4ISO/IEC Directives Part 2 prescriptive language applied correctly throughout. SRA "should" guidance characterised as a recommendation in Section 3 pull quote paraphrase. Pull quote replaced with paraphrase and attribution — original SRA text not reproduced verbatim. Part E procedure obligations use "shall" correctly.PASS
- G5REG-AIMS-SUP-001 download linked ungated in Section 6 doc-cta block. All cross-referenced document IDs present and correct: REG-AIMS-SYS-001, REG-AIMS-DATA-001, PROC-AIMS-LIFE-001 (T8 trigger), REG-AIMS-RISK-001, POL-AIMS-001, DOC-AIMS-DISC-001, PROC-AIMS-HITL-001, RACI-AIMS-001.PASS
- G6Footer includes predecessor link (← G8 — Human Oversight). Series-final article: no successor link required per series convention. Breadcrumb corrected to /legal/ path structure. Section 7 AIMS document index includes all nine documents with ISO clause references.PASS
- G7Standalone disclaimer section present between final-cta and footer. Includes: educational purpose notice, fictional scenario notice for Halstead & Cole LLP, Priya Anand, Pemberton Capital LLP, and Thornbury Legal Ltd., document reference ART-AIMS-G009 v1.1, publication date, next review date, retention period, and canonical URL.PASS
- G8No inline policy or template content requiring a document-control block. REG-AIMS-SUP-001 is described and linked ungated. Gate marked N/A.PASS
- G9H1 includes primary keyword "supply chain" in first 10 words of the title question. Stat bar appears within first screenful. Schema.org Article JSON-LD with isPartOf series reference present in <head>. Canonical URL set: https://unuslondon.com/legal/iso-42001-gap-9-supply-chain. Meta keywords present (6 terms including Cl. 8.5 reference and "law firm").PASS
- G10IEC 82079-1:2012 notice hierarchy CSS comment block present above :root. .notice-danger class absent from stylesheet — removed entirely. DANGER not used anywhere in body or CSS. WARNING used for self-hosted governance exposure (Section 1) — regulatory/conduct consequence. CAUTION used for unverifiable vendor claims (Section 5) — commercial/procedural risk. NOTE used for Cl. 8.5 / Cl. 8.3 complementarity (Section 2). ITIL 4 NOTE present with indigo styling, mapped to Supplier Management and Risk Management (Section 2). sn-risk supply map node corrected from danger red to warning amber palette — ungoverned supply chain is an operational risk, not a physical hazard. Legend entry corrected to match.PASS