Our current instrument:
Management Information Integrity Framework
v2026.2 | Published by Reality & Reason
The Problem
Boards depend on information that is accurate, complete, and fairly presented. Where AI is used in the preparation of Management Information, failure modes arise that conventional reporting controls were not designed to detect.
This instrument does not address whether AI systems are fair, ethical, or technically sound. It addresses whether AI's reasoning can reach the board without being identified and examined along the way.
The risk is not AI use. The risk is undisclosed AI influence on decision-critical narrative, weak attribution, and absent provenance, such that the board cannot tell whether it is relying on human judgement, model outputs, or AI-shaped framing.
The Framework identifies two failure signatures, drawn from the MKAI Structural Dynamics research programme:
Residual Logic.
AI's patterns, assumptions, and framing persist in the final output even after human editing. The draft shaped the document. The structure survives. The fingerprints remain.
Absent Equals.
Equally valid options were eliminated before the board saw them. The board chose from a narrowed field without knowing alternatives existed. The presented option was correct, but so were the options that were never shown.
What the Framework Contains
Four Board Motions
Each drafted as a Board resolution, copy-paste ready for adoption:
Three Governance Postures
Control requirements calibrated per document stream:
Standard
Default controls for routine governance documents. Single attestor, traceability record, basic validation.
Conservative
Enhanced assurance. Dual attestor sign-off, option-set disclosure, enhanced traceability.
Strict
Independent validation, adversarial challenge, source separation, full transparency to the board. Required for financial results, risk appetite statements, regulatory submissions, and other high-stakes document streams.
Three Behavioural Controls
New in v2026.2
Designed to change the authoring process upstream rather than detect invisible phenomena after the fact:
Option-Set Disclosure.
Requires documentation of alternatives considered, or an Option Limitation Statement explaining why alternatives were not developed. Makes narrowing visible.
Adversarial Challenge Record.
For Strict posture documents, a structured counter-factual prompt requiring AI to generate arguments against the recommendation and identify unconsidered alternatives.
Source Separation.
For Strict posture documents, Executive Summary and Recommendation sections must be human-drafted. AI may inform but must not generate the persuasive core.
What You Receive
| Document | Description |
|---|---|
| Management Information Integrity Framework v2026.2 | Core governance instrument. Definitions, Board Motions, Minimum Controls, Conformance Reference, Licence and Adoption. PDF and Word. |
| Conformance Test v2026.2 | 15 pass/fail items and 10 maturity indicators for self-assessment or Internal Audit use. |
| Crosswalk v2026.2 | Mapping to NIST AI RMF and ISO/IEC 42001, including behavioural controls summary. |
| How to Table This | One-page adoption guide. Before the meeting, at the meeting, after the meeting (90-day checklist). |
| Traceability Report (Worked Example) | The traceability record for the Framework itself, demonstrating the process the instrument requires. |
| Audit Committee Briefing v2026.2 | Step-by-step briefing for the Audit Committee's first oversight session. Covers the Committee's mandate, the 90-day checklist, ongoing responsibilities, and what evidence to expect. |
The Framework is supplied in both PDF (reference version) and Word (editable working version) to support internal governance adoption. All other supporting documents remain PDF.
Sample Content
From Section 5.3 — Human Attestor Attestation
Human Attestor Attestation
This document contains AI-Assisted Content as defined in the Management Information Integrity Framework.
I confirm that:
- AI-Assisted Content has been identified and disclosed as required;
- applicable Traceability Record requirements have been satisfied;
- validation steps commensurate with the applicable Governance Posture have been completed;
- Option-Set Disclosure or Option Limitation Statement requirements have been satisfied [Conservative/Strict only]; and
- the content is accurate and complete to the best of my knowledge.
Human Attestor: _________________________ Date: _____________
Sample Conformance Test Item
| # | Requirement | Source | Evidence |
|---|---|---|---|
| 1 | Board resolution adopting the Framework is recorded in minutes with effective date | Motion 1 | Board minutes |
Who This Is For
Company Secretaries, Heads of Internal Audit, Chief Risk Officers, General Counsel, and Audit Committee Chairs in listed companies and large unlisted businesses operating under or aligned with the UK Corporate Governance Code, where AI is used in the preparation of board papers, executive committee papers, risk reports, or other Management Information.
Adoption Path
The Framework is designed to be adopted by Board resolution and implemented through existing governance structures. No external support is required.
By day 90 after adoption, three things must exist: Governance Postures recorded in the Board-Critical Narrative Register, attestation block embedded in board paper templates, and initial conformance assessment completed and reported to the Audit Committee.
What Happens After Adoption?
The bundle includes everything needed from Board resolution through to the first Audit Committee oversight cycle.
The How to Table This guide covers the Board meeting: what to circulate beforehand, what motions to pass, and what to record in the minutes.
The Audit Committee Briefing covers the Committee's first oversight session: their mandate under the Framework, the evidence they should expect, and their ongoing responsibilities.
The Conformance Test provides the 90-day assessment: 15 pass/fail items and 11 maturity indicators, with clear outcomes (Pass, Conditional, Fail).
No external support is required at any stage.
Relationship to Other Frameworks
The Framework is supplementary to existing governance, risk, and control frameworks. It does not replace financial reporting controls, internal control frameworks, risk management frameworks, model risk management frameworks, or data governance frameworks. A separate Crosswalk maps requirements to NIST AI RMF and ISO/IEC 42001.
Alignment with the UK Corporate Governance Code
Provision 29 of the UK Corporate Governance Code applies to financial years beginning on or after 1 January 2026. For those reporting periods, Provision 29 expects boards to include, in the annual report, a declaration on the effectiveness of material controls as at the balance sheet date, covering material financial, operational, reporting, and compliance controls. The first declarations will generally appear in annual reports covering 2026 reporting periods, published in 2027 for calendar-year companies.
Where AI is used in the preparation of Management Information that informs board decisions, the integrity of that information is a control concern within the Provision 29 perimeter. The Management Information Integrity Framework is designed to be referenced in the board's risk and control documentation and to provide the evidence trail that supports the Provision 29 declaration in respect of AI-Assisted Content.
The Framework does not replace the board's existing risk management framework or internal control framework. It supplements them with controls specific to AI-mediated reporting, calibrated to the Governance Posture applied to each document stream.
Executive Summary
Download the Executive Summary (PDF, 3 pages) to circulate before recommending the spend.
Download Executive SummaryPublished by Reality & Reason, an independent editorial body. No consulting, no implementation, no vendor affiliations.
Pricing
| Licence | Covers | Price |
|---|---|---|
| Single Entity | One legal entity, full document bundle | £2,500 + VAT |
| Group | Parent entity plus subsidiaries | £7,500 + VAT |
Registered charities and public sector bodies: 50% of applicable rate.
Future editions may be adopted at the renewal price applicable at the time of publication (currently £1,500 Single Entity, £4,500 Group). All minor updates within the edition year are included at no additional cost. Purchases made within 90 days of a major edition release include the incoming edition at no additional cost. Renewal is optional. Prior editions remain valid until superseded.
Licences are issued to a named organisation or group and the supplied documents are identified accordingly. Each licensed bundle includes organisation-specific identification for traceability and audit purposes.
Redistribution outside the licensed organisational scope is not permitted. If use extends beyond the licensed entity or group, the appropriate licence will be required.
Request Invoice
Adoption model
This licence provides the document bundle for internal governance adoption and use. It does not include implementation services, legal advice, certification, or managed onboarding. This keeps the focus on the instrument itself while allowing organisations to adopt it through their own governance processes.
Frequently Asked Questions
Related Analysis
The following published work examines the conditions the Framework addresses:
The Governance Contradiction — examines what happens when a governance framework contains the evidence against its own model.
fosterfletcher.com
The Signature Fiction — examines what happens when AI drafts the reasoning and a human signs it off.
fosterfletcher.com
Residual Logic — examines what happens when AI-generated framing persists through every subsequent edit.
fosterfletcher.com
The Big Four AI Audit Tool Register — what Deloitte, PwC, EY, and KPMG publicly disclose about AI in audit, mapped against FRC findings. Reality & Reason, May 2026.
realityandreason.org/audit
The First Annual Reports of the LLM Era — a study of language drift in 150 SEC 10-K filings across 2019, 2022, and 2024. SSRN, 2026.
papers.ssrn.com
The Subtraction Study: Measuring Capability Reduction in Enterprise AI Configurations — tests what enterprise governance configurations remove from large language model output. SSRN, May 2026. Publication pending.
papers.ssrn.com
The research behind the Framework, including the Corporate Disclosure Prose Drift study, is published through MKAI at mkai.org.