UNUS London Official Governance Academy ISO/IEC 42001:2023 Module 08
Course Progress
08 / 10
Module 08 — AI Client Disclosure & Course Continuation

Your Client Has a Right to Know.
Halstead & Cole Now Has a Way to Tell Them.

The final module of the ISO 42001 Legal Track. Two obligations converge here: ISO 42001 Annex A.2 (transparency to affected parties) and the SRA's expectation that solicitors consider disclosure before using AI in client matters. This module builds the three-template disclosure pack that closes Gap 7 — and gives Halstead & Cole a complete answer to every question on Pemberton Capital's panel questionnaire.

ISO/IEC 42001 Annex A.2 SRA Code of Conduct DOC-AIMS-DISC-001 Module 08 · Disclosure Pack
0 Law firms in baseline assessments with a documented AI disclosure procedure and template
3 Disclosure trigger scenarios applicable to Halstead & Cole — none currently documented
Q5 Pemberton Capital questionnaire — the question this module closes
5/5 Pemberton Capital questions answerable by end of this module
Section 01 — The Scenario

The Question Clients Are Starting to Ask

"Was AI used in preparing my case?" It is a question that more clients are asking. It is a question that panel procurement teams are putting in writing. It is a question that Ayinde v Haringey [2025] EWHC 1383 implicitly raised for every firm whose fee earners use AI tools in client work. For most law firms, it is a question they cannot answer with a documented process, a consistent approach, or a ready-made disclosure.

Firm
Halstead & Cole LLP
AIMS foundation
Policy · Register · Risk · Data — ALL ACTIVE
Pemberton status
Q1–Q4 answered · Q5 still pending
Care letter today
No AI disclosure clause
Matter-file records
Zero documented disclosure decisions across active matters
What changes today
Three templates · one decision framework · one matter-file record
Module number
08 · Disclosure Pack
Outcome
5 of 5 Pemberton questions answered with documented evidence

"Solicitors should consider whether clients should be told when AI has been used in their matter, having regard to the nature of the AI use and the client's reasonable expectations."

— SRA · Compliance Tips for Solicitors Regarding the Use of AI and Technology · 9 February 2026

The undisclosed use scenario

If a client later discovers that AI was used in their matter — through a subject access request, a court disclosure exercise, or a complaint investigation — and the firm has no record of having considered whether to disclose that use, the firm cannot demonstrate that the omission was a deliberate, considered decision rather than an oversight. The SRA's investigation will focus on whether the COLP had a process for making that decision. The absence of DOC-AIMS-DISC-001 is the absence of that process.

Section 02 — What Annex A.2 Requires

The Operative Word Is "Should" — and Why That Matters

ISO 42001 Annex A.2 uses the word should — not shall. In ISO/IEC Directives Part 2, "should" is a recommendation rather than a hard requirement. This is a deliberate calibration: blanket mandatory disclosure of every AI tool used in every client interaction would be operationally impractical and contextually inappropriate. The standard recognises this. What the standard does require is that the firm have a documented process for making the disclosure decision.

Plain English summary

Annex A.2 does not prescribe the answer to "should you tell the client?" — it requires that the firm ask the question, consider the answer, and record the decision in the matter file. DOC-AIMS-DISC-001 is the instrument that captures that question, that consideration, and that decision. The disclosure templates are the tools for how the answer is communicated — but the procedural discipline is the core compliance obligation.

"Should" in ISO 42001 is not "optional" in SRA terms

ISO 42001's should recommendation does not displace the SRA's conduct obligations. The SRA's 9 February 2026 guidance creates an obligation to consider disclosure as part of the duty to act with integrity and in the client's best interests. That obligation flows from SRA Principles 2, 4, and 7 — not from Annex A.2. ISO 42001 provides the structure; the SRA provides the professional obligation. DOC-AIMS-DISC-001 satisfies both simultaneously.

The two-part test — nature of use, reasonable expectations

The SRA's formulation creates a two-part test for the disclosure decision: (1) the nature of the AI use, and (2) the client's reasonable expectations. Both are context-dependent. A spell-checker does not require disclosure in the same way an AI-drafted chronology does; a sophisticated corporate client in regulatory matters expects disclosure that a one-off consumer client may not anticipate. This is precisely why a documented procedure — rather than a blanket rule either way — is the right instrument.

The K&L Gates ISO 42001 certification of 9 March 2026 — the first law firm globally — publicly demonstrated that a large law firm had satisfied the transparency obligations of Annex A.2. Panel clients comparing firms on AI governance credentials will increasingly treat the absence of a documented disclosure procedure not as a neutral omission but as a governance gap.

— Gap 7 Article · Pemberton Capital Implications
Section 03 — The Disclosure Decision Framework

Walk Through the Five-Question Decision Tree

The framework below is the operational heart of DOC-AIMS-DISC-001. It applies to every AI use in every matter. Click YES or NO at each step to see what action follows. The framework produces either a disclosure (Template A, B, or C) or a recorded non-disclosure decision — both defensible in a regulatory review.

AI Client Disclosure Decision Tree · DOC-AIMS-DISC-001 · Annex A.2 · Click YES / NO at each step
Q1
Was AI used to generate, draft, summarise, or materially inform any substantive content — advice, correspondence, submissions, or analysis — that will be delivered to or relied upon by the client?

Both outcomes are defensible — provided they are recorded

The SRA's expectation is not "disclose every AI use" — it is "consider disclosure in every matter and record the decision." A firm that records a considered decision not to disclose minor AI use (Q4 YES path) has discharged the obligation. A firm that records a decision to disclose (Q3 YES path or Template B issued) has discharged it differently. The audit failure is not the disclosure decision itself — it is the absence of the recorded decision. The matter-file AI Disclosure Record is what makes either outcome defensible.

Section 04 — Failure Modes

Four Ways a Disclosure Process Fails

Most firms do not fail because they actively decide not to disclose AI use — they fail because they have no decision framework at all. The four non-conformance categories below cover the realistic failure modes for Annex A.2 compliance. Click each one to see why it fails.

Failure modes reviewed
0 / 4

The firm has no framework for deciding whether to disclose AI use. Individual fee earners make ad hoc judgments that are inconsistent across the practice. Neither outcome — disclosure or non-disclosure — is recorded in the matter file. When an SRA investigation or panel-client review asks the question, the firm cannot produce a documented answer for any specific matter.

Audit finding: "No documented AI client disclosure procedure exists. Annex A.2 transparency expectations unaddressed; SRA February 2026 guidance obligations unevidenced."

The standard engagement letter has not been updated to include an AI disclosure clause. New matters opened after AI tools were adopted contain no reference to AI use — even where AI has been used to draft correspondence on that very matter. Template A from DOC-AIMS-DISC-001 addresses this with a single insertion into the standard care letter; it takes one working session and immediately puts all new matters onto the correct footing.

Audit finding: "Standard client care letter template not updated to address AI use. Annex A.2 prospective disclosure obligation unaddressed for matters opened after the AI policy was issued."

Even where disclosure has been made informally — verbally or in passing correspondence — it has not been documented. The firm cannot demonstrate to the SRA, a panel client, or in litigation that it considered and made a deliberate decision in any specific matter. The matter-file AI Disclosure Record is the one-page record that closes this — capturing the AI systems used, the Q1–Q5 decision tree outcome, the disclosure action taken, and the fee earner sign-off.

Audit finding: "No matter-level record of AI disclosure decisions available for any active matter. SRA expectation of professional judgement under February 2026 guidance unevidenced."

Panel clients like Pemberton Capital have explicit AI governance questionnaires. Treating them the same as individual consumer clients — with the same care letter disclosure clause — fails to address the institutional client's need for a structured, documented account of the firm's AI governance framework. Template C is the institutional client AI governance statement — the document that answers Pemberton Capital Question 5 with the specificity a panel client requires.

Audit finding: "No template or framework for responding to panel-client AI governance questionnaires. Institutional client disclosure obligations addressed only by generic care letter clause."

Section 05 — The Fix

DOC-AIMS-DISC-001 · Three Templates for Three Contexts

DOC-AIMS-DISC-001 is not a policy document — it is a template pack. It contains three ready-to-use disclosure instruments, the decision framework, and the matter-file recording mechanism. The three templates map to three disclosure contexts: prospective (Template A), matter-specific (Template B), institutional (Template C).

01
Template A · Prospective

Care letter / engagement agreement AI clause

A standard clause for insertion into the firm's client care letter template. It discloses that the firm may use AI tools, describes the governance framework, confirms human oversight applies, and invites the client to raise concerns. General and prospective — covers AI use across the matter rather than any specific instance. Operative for all AI use in that matter unless a specific HIGH-risk use triggers Template B additionally.

02
Template B · Matter-specific

Standalone AI disclosure letter

A standalone letter for matters where Template A was not in the original care letter and AI has been used; where a HIGH-risk AI system has been used; or where professional judgement indicates disclosure beyond Template A. Specific — identifies the AI system (at category level), the task performed, the human oversight applied, and the client's right to ask further questions.

03
Template C · Institutional

Panel and institutional client governance statement

A structured response document for panel questionnaires and institutional client AI governance enquiries. Addresses the firm's AI governance framework — the AIMS, the gap register, the ISO 42001 alignment, the disclosure procedure. The document that answers Pemberton Capital Question 5.

Template A — client care letter AI clause
Template A · Client Care Letter AI Clause Insert into standard care letter · DOC-AIMS-DISC-001

Use of artificial intelligence

We may use artificial intelligence tools to assist in providing legal services on your matter. Where we do so, such tools may be used to assist with tasks including drafting, research, document review, or summarisation. All AI-assisted work is reviewed by a qualified member of our team before it is used in your matter, and we remain fully responsible for the quality and accuracy of all work we deliver to you.

We maintain a governance framework for AI use, including documented oversight procedures and data protection controls. We do not use AI tools that process your personal information in ways that are not covered by appropriate data protection agreements, and your confidential information remains subject to our professional duty of confidentiality at all times.

If you have any questions about our use of AI tools in your matter, or if you would prefer that we do not use AI tools in any specific aspect of your matter, please let us know and we will accommodate your request wherever possible.

Template B — matter-specific disclosure letter
Template B · Matter-Specific AI Disclosure Letter Issue as standalone letter or email · Record in matter file

Dear [Client name],

Re: [Matter reference] — Use of AI assistance in your matter

I am writing to let you know that, in the course of working on your matter, we have used an AI tool to assist with [brief description of task — e.g. drafting the chronology / reviewing the documents / preparing the first draft of correspondence].

The AI-generated content was reviewed by [fee earner name and role] before it was used, and any errors or inaccuracies were corrected. We are satisfied that the work delivered to you accurately reflects our professional advice and is not solely the product of AI generation.

If you have any concerns about this, or if you would like us to explain how we use AI tools in more detail, please do not hesitate to contact me.

Template C — institutional governance statement
Template C · Institutional / Panel Client AI Governance Statement Structured response · For panel questionnaires · DOC-AIMS-DISC-001

[Firm name] operates an Artificial Intelligence Management System (AIMS) aligned with ISO/IEC 42001:2023. Our current documented controls cover:

AI Policy (POL-AIMS-001) · leadership-approved, six mandatory sections
AI System Register (REG-AIMS-SYS-001) · every deployed AI system, with risk tier and oversight mechanism
Risk Register (REG-AIMS-RISK-001) · eight-factor structured AI risk assessment for each system
Data Governance Register (REG-AIMS-DATA-001) · lawful basis, DPA, overseas transfer, data minimisation
AI Disclosure Pack (DOC-AIMS-DISC-001) · three templates, decision framework, matter-file records
Human Oversight Procedure (PROC-AIMS-HITL-001) · Module 06 preview

Client disclosure obligations are addressed through a structured decision framework. For any matter in which AI is used to generate substantive content, our COLP-signed procedure determines whether and how disclosure is made — and the decision is recorded in the matter file. We welcome the opportunity to provide further detail on any specific AI system or governance control.

What NOT to disclose in Template B

Template B should not identify the specific AI vendor by name (OpenAI, Microsoft, Anthropic, etc.) in client correspondence, unless the client specifically asks. Naming vendors can create client anxiety disproportionate to the actual risk, and may expose the firm to follow-up questions about vendor security the fee earner cannot answer authoritatively on the spot. The template uses category language ("an AI drafting tool") rather than product language. Vendor details live in REG-AIMS-DATA-001 and REG-AIMS-SYS-001 — available on request, but not proactively disclosed in the standard letter.

Section 06 — The Five Pemberton Questions, Now Answerable in Full

Halstead & Cole's Complete Questionnaire Response

With Module 08 complete, every question on Pemberton Capital's five-question AI governance questionnaire can be answered with specific, documented evidence from the AIMS. The summary below is the firm's response capability — every question maps to a controlled document, signed by senior leadership, updated on the documented review cadence.

Pemberton Capital AI Governance Questionnaire · Response Capability Map · 30-day deadline
Q1
"Who owns AI risk in your firm?"
→ RACI-AIMS-001 (Module 03 · Gap 2). AI Risk Owner and AI Governance Lead formally designated by management resolution. COLP holds regulatory interface. Documented names, authorities, and accountabilities — not a single point of failure.
Q2
"What AI tools does your firm use in client matters?"
→ REG-AIMS-SYS-001 (Module 04 · Gap 3). Every AI system registered with risk tier, intended use, oversight mechanism, vendor, and data categories. M365 Copilot split by deployment context (Word, Outlook) per system-level documentation requirements. Sys-005 blocked pending vendor assessment.
Q3
"How do you identify and manage AI-specific risks?"
→ REG-AIMS-RISK-001 (Module 05 · Gap 4). Eight-factor structured AI risk assessment using a 5×5 likelihood × impact matrix. Two High-rated risks documented (transparency, oversight adequacy) with target remediation dates. Ongoing quarterly review.
Q4
"How do you monitor AI systems after deployment — and how do you protect client data in them?"
→ REG-AIMS-DATA-001 (Module 07 · Gap 6). Lawful basis per system and data category, vendor DPA confirmed, overseas transfer mechanism recorded, data minimisation rules per system, special-category data blocked until DPIA in place. UK GDPR + ISO 42001 Cl. 8.3 satisfied simultaneously.
Q5
"How do you communicate AI use to clients?"
→ DOC-AIMS-DISC-001 (Module 08 · Gap 7). Five-question disclosure decision framework. Three templates: Template A (care letter clause), Template B (matter-specific letter), Template C (institutional panel-client governance statement). Matter-file record for every disclosure decision.

Five questions · five documents · one framework

Halstead & Cole has gone from "we don't have an AI policy" (Module 02 opening) to a documented, leadership-approved AIMS that answers every question on a £620,000/year panel client's governance questionnaire with specific, evidenced instruments. The framework is defensible under both ISO 42001 audit and SRA regulatory scrutiny — and it is reproducible in any mid-sized UK law firm without disproportionate investment. The remaining operational control (G8, Human Oversight Procedure, PROC-AIMS-HITL-001) is the natural next module in any continuation programme.

Quick check — test your understanding
Question 1 of 3 — Why is "should" in Annex A.2 different from "should" in casual English?
Correct Answer: B Per ISO/IEC Directives Part 2: "shall" denotes a requirement; "should" denotes a recommendation; "may" denotes a permission. A "should" requirement is one the standard expects the firm to follow unless there is a documented justification for departing from it — far stronger than casual English usage suggests. Combined with the SRA's professional judgement obligation under its February 2026 guidance, "should" disclosure in ISO 42001 becomes operationally indistinguishable from a binding requirement.
Question 2 of 3 — A fee earner uses ChatGPT (free tier, personal account) to draft an email to a client. What is the correct procedure?
Correct Answer: D POL-AIMS-001 Section 3 (Module 02) explicitly prohibits the use of personal or consumer-grade AI accounts for matter-related work. The use of free-tier ChatGPT for a client email is a policy breach in its own right — regardless of whether disclosure is later required. The COLP must be notified, the matter re-assessed (any unauthorised disclosure has occurred), and the AI System Register updated. The disclosure framework does not operate until the policy compliance issue has been resolved.
Question 3 of 3 — Why is recording the non-disclosure decision (Q4 YES path) as important as recording a disclosure?
Correct Answer: C The SRA's expectation under the 9 February 2026 guidance is that disclosure is considered and the decision recorded — not that disclosure itself must happen. A firm that records a considered decision not to disclose minor AI use (e.g. spell-check) has discharged the obligation just as effectively as a firm that records a decision to disclose. The audit failure is the absence of the recorded decision, not the disclosure outcome itself. Both paths of the decision tree are defensible — provided the matter-file record exists.
Section 07 — Course Completion Gate

Module 08 Completion Checklist

The final completion checklist for this phase of the AIMS. Ticking every box below signals that Halstead & Cole has the documented, defensible AI governance framework needed to respond to Pemberton Capital in full — and to satisfy both ISO 42001 audit and SRA regulatory expectations across all ten modules of this course.

Checklist progress
0 / 16
Understanding the Scenario
I can explain why "we take data protection seriously" is not an answer to the SRA's confidentiality and disclosure obligations
I can describe what the Ayinde chain of causation implied for firms that used AI tools without disclosure procedures
I understand why both disclosure and non-disclosure are defensible outcomes — provided the decision is recorded
I can describe the strategic context of K&L Gates' ISO 42001 certification and what it signals for panel-client expectations
Understanding Annex A.2
I can explain why "should" in Annex A.2 is operationally binding even though it isn't a "shall" requirement
I can describe the SRA's two-part test: nature of AI use, and client's reasonable expectations
I understand why Annex A.2 does not mandate blanket disclosure but requires a documented decision framework
I can state what Pemberton Capital Question 5 specifically probes and how Template C answers it
Tree & Failure Modes
I can describe all four NF categories and the audit finding each one would generate
I can walk through the five-question disclosure decision tree and identify each branch outcome
I understand why Template A is prospective (general, covers the matter) while Template B is specific (this use, in this matter)
I can explain why vendor names should not appear in Template B by default — and when they should
Readiness to Respond
I can describe all five Pemberton questionnaire answers and the documents that evidence each one
I understand the matter-file AI Disclosure Record requirement and the fee earner sign-off obligation
I can brief the firm's fee earners on the decision tree and on the Template A insertion into the care letter
I am ready to submit the complete questionnaire response with the five controlled documents attached