Why AI Transparency Is Now a Board-Level Issue

By Karl George MBE  |  GovernAI Academy  |  old.governanceai.io

Here is an uncomfortable truth for every board in the country: your people are already using AI, and most of them are not telling you about it.

They are using it to draft emails, summarise reports, analyse data, prepare board papers, and generate client advice. Some of these tools have been approved. Many have not. And in the vast majority of cases, nobody is disclosing the AI contribution or verifying the output.

This is not a technology problem. It is a governance problem. And it is one that has already cost careers, reputations, and millions of pounds.

The consequences are already here

In January 2026, the Chief Constable of West Midlands Police retired with immediate effect after it emerged that his force had used Microsoft Copilot to generate intelligence for a safety report, and that the AI had fabricated a reference to a football match that never took place. The Chief Constable initially denied using AI. The police watchdog found confirmation bias and exaggerated intelligence. The Home Secretary publicly declared she had lost confidence in him.

In 2025, Deloitte was forced to refund nearly $300,000 to the Australian government after a researcher discovered that a 237-page report was littered with fabricated academic citations, non-existent journal articles, and a made-up quotation attributed to a federal court judge. All hallmarks of generative AI. Deloitte had not disclosed the use of AI in the original report. Within months, a second Deloitte report for the Canadian government, costing C$1.6 million, was found to contain identical problems.

In the legal profession, over 600 documented cases worldwide have now been identified where lawyers submitted court filings containing AI-fabricated citations. Courts have imposed fines, referred lawyers to disciplinary bodies, and dismissed cases entirely.

The pattern is always the same

AI was used to produce or substantially contribute to a work product. The AI contribution was not disclosed. The output was not verified. The error was discovered externally. And in every case, the denial or concealment was more damaging than the original mistake.

Regulation is catching up

The EU AI Act becomes fully operational in August 2026. Article 4 requires organisations to ensure AI literacy among all personnel. Article 50 imposes transparency obligations including requirements to inform users when they interact with AI. The fines are substantial: up to €35 million or 7% of global turnover.

Boards that cannot answer the question, “How is AI being used in this organisation, and what governance is in place?” are carrying hidden risk that is about to become visible.

What boards need to do

AI transparency is not about banning AI. It is about governing it. Organisations need a simple, proportionate framework that classifies the level of AI involvement in work products, applies governance that is proportionate to risk, ensures external disclosure to clients and stakeholders, and provides the board with evidence that AI is being used responsibly.

I have published the AI Transparency Framework™ through the GovernAI Academy. A free summary deployment guide is available at governaiacademy.com, giving you the Index, the risk matrix, the self-assessment process, and template disclosures to get started immediately. The full workbook, with detailed case studies, worked examples, and a complete implementation guide, is available to organisations that engage with The Governance Forum through our workshops and consultancy services.

Because the question is no longer whether your organisation uses AI. It is whether you can prove you are governing it.

#AIGovernance #BoardLeadership #AITransparency #GovernAI #GovernAIAcademy