Pensions Regulator: trustees should embrace AI but govern it carefully

TL;DR:

• The Pensions Regulator’s CEO Nausicaa Delfas has issued the most detailed TPR statement on AI to date, framing AI as an opportunity for trustees but flagging governance, transparency and accountability as the conditions for safe adoption.
• The 6 May “Looking Ahead to 2036” speech is consistent with TPR’s June 2025 directive that trustees “must understand the role of AI in industry” but goes further by spelling out the operational risks: data quality, bias, explainability and security.
• Legal analysis from Burges Salmon places the speech alongside PASA’s October 2025 AI guidance and suggests trustees should now actively consider whether an explicit AI policy is appropriate at scheme level.

The framing matters because TPR is the second UK financial-services regulator after the FCA to put AI on its top-tier agenda, and the first to do so for the trustee community specifically. UK occupational pension schemes hold over £2 trillion in assets and serve more than 27 million members — the regulatory perimeter affected is wider than the equivalent FCA-supervised AI guidance.

What TPR is actually asking trustees to do

Delfas’s speech sets out three operational threads. First, AI can improve regulatory effectiveness — identifying risks earlier, targeting interventions more precisely, supporting faster decisions on member outcomes. Second, AI deployment must be governed with strong oversight, transparency and accountability. Third, AI should support human judgement, not replace it, particularly where statutory obligations and public expectations apply.

Burges Salmon’s pensions team — Callum Duckmanton and Chris Brown — frame the practical implication as a trustee-board agenda item: AI governance review, member-risk mitigations (particularly around misinformation produced by consumer AI tools), and engagement with third-party providers on AI use in administration and asset management.

UK angle: a regulator catching up with adoption

The speech aligns with what pensions practitioners have been seeing on the ground. AI moved from sidepiece to centre-stage on trustee board agendas in 2025, and TPR’s statement is closer to validation of existing trustee concern than to a directive. The risks Delfas highlighted — bias, explainability, security, data quality — are the same risks the Logicalis CIO survey published this week showed UK enterprise leaders broadly unprepared for. For pension schemes, the additional dimension is fiduciary: trustees carry personal duty of care obligations that ordinary corporate AI deployment does not face.

Looking forward

Expect TPR’s General Code review later this year to incorporate AI-specific expectations for the first time, and for PASA to issue an updated AI guidance document responding to Delfas’s speech. The bigger story is the regulatory pattern: TPR joining FCA, MHRA, Ofcom and the ICO in publishing AI guidance, even before the Regulating for Growth Bill set out in this week’s King’s Speech reaches Parliament. The UK’s sectoral-regulator AI model is producing visible coordination — if not always identical positions — across the prudential and conduct perimeter.