Pensions regulator gives trustees ten questions to ask about AI

The Pensions Regulator has updated its guidance on AI use by pension schemes and their service providers, setting out ten questions trustees should be able to answer for each significant use. The point is practical: trustees remain accountable, even when a provider supplies the technology.

Watch Desk analysis

What happened

The regulator says trustees should know what AI is being used and why, what process or member outcome it improves, and which decisions it may influence. They should also be able to explain what data the system uses and how it is handled, who is accountable, how the system is tested, and how users or members can challenge a decision. The remaining questions cover human overrides, what happens if the AI fails or becomes unavailable, and how its use has benefited members.

The guidance applies to significant AI use both inside schemes and at service providers. The regulator says trustees are not expected to become AI experts, but they should understand where it is used and be able to challenge providers. Read the guidance as described by Pensions Expert.

Why it matters

Pension schemes can rely on AI without using it directly themselves: a provider’s system may still affect processes or decisions that matter to members. The checklist gives trustees a way to ask about those uses without pretending that every board needs to become a model-development team.

It also makes the accountability line hard to miss. Outsourcing a system does not outsource the duty to understand how it is used, test its outputs or respond when it fails.

Our read

This is useful governance guidance because it turns “keep an eye on AI” into questions trustees can put to their own teams and suppliers. The test is whether those questions lead to clear answers, meaningful challenge and a workable plan when the system is unavailable, rather than another tidy document in the governance folder.

What to watch

• How schemes apply the questions to significant uses by their service providers.
• Whether trustees can explain how AI use benefits members, not just internal processes.
• How schemes test outputs, handle challenges and maintain processes when systems fail.

Discussion spark: Should trustees be expected to answer these questions for every significant AI use by a provider, or is that asking too much of boards that do not control the system?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.