Board oversight of AI means being able to demonstrate four things: the board knows what AI the company uses, understands how it is classified under the EU AI Act, has assigned ownership of the resulting obligations, and reviews the position on a regular cycle — with all four minuted. That bar is not yet being met in most Irish boardrooms: the Institute of Directors found in Q1 2026 that 65% of Irish directors don’t understand the new AI rules.
Why AI is now a board matter
Three forces have moved AI from an IT decision to a governance one. The EU AI Act attaches staged legal obligations — and penalties — to how a company uses AI, with the earliest duties in force since February 2025. Ireland is building national enforcement through the Regulation of AI Bill 2026 and the AI Office of Ireland. And counterparties are moving faster than regulators: customers, insurers and funders increasingly ask for evidence of AI governance as a condition of doing business. Directors’ ordinary duties of care and oversight now have an AI-shaped section, whether the agenda reflects it or not.
The four things a board must be able to show
- Visibility: a current AI register listing every system in use, including AI features inside ordinary software. A board that cannot answer “what AI do we use?” has no oversight to demonstrate.
- Classification: each system assessed against the Act — prohibited, high-risk, transparency or minimal — with the reasoning recorded and the recruitment-tool trap checked explicitly.
- Ownership: a named executive owner for AI compliance overall and for each system, so that obligations and deadlines have a person attached, not a department.
- Review: AI on the board agenda at a set frequency, with the register, the deadline position and any incidents reported — and the discussion minuted.
Directors’ own literacy duty
The Article 4 AI literacy duty scales with role and responsibility, and the people responsible for overseeing an organisation’s AI use sit at the top of that scale. A board does not need to understand transformer architectures. It needs to understand the Act’s risk categories, the company’s own classifications, the deadlines, and the questions to ask management — a half-day briefing, refreshed annually, with attendance recorded. The IoD’s 65% figure suggests this single step would put a board ahead of most of its peers.
The questions to ask management at the next meeting
- Do we have a complete register of the AI in use, and when was it last reviewed?
- Do any of our systems touch the high-risk categories — recruitment and worker management above all?
- Are we ready for chatbot disclosure on 2 August 2026, and are we tracking the Annex III date — 2 December 2027 under the Omnibus agreement, subject to formal adoption?
- Can we evidence Article 4 literacy training for staff — and for this board?
- Who owns AI compliance, and what would they show a regulator or major customer tomorrow?
Oversight you can hand across the table
AI Register Ireland was built for exactly this evidence problem: a register the board can actually read, classifications with reasoning, deadlines tracked, and a one-page position for the board pack. See pricing.