The Pensions Regulator’s 29 October guidance asks trustees to assess AI that arrives through existing software and updates. A supplier review at the start of a contract cannot cover every later change in what the system does.[1]
The guidance explains existing duties. Trustees remain accountable for outsourced work and should seek assurance relevant to their scheme; a generic provider statement may be insufficient where members could be affected.[1]
Usage is easier to see than oversight
A survey published by Aon on 16 October covered 61 UK defined benefit schemes. It found that 76% of respondents reported some trustee AI use. Yet 61% did not know how their advisers used it, and 73% carried out no specific assessment of providers’ AI governance.[2]
These are findings from the participating schemes. The announcement provides neither a population sampling frame nor a response rate that would support treating them as national prevalence estimates.[2]
One high-risk example in the guidance is a member-service system using sensitive information that can change records or initiate transactions. Trustees should know how people can challenge decisions and override the system. They should also have processes to detect member harm, correct errors and provide appropriate redress.[1]
Where assurance falls short, trustees should consider further enquiries, escalation or stronger controls, and revisit contractual terms when agreements are made, renewed or changed.[1]
Sources & further reading
- Use of artificial intelligence in pension schemes The Pensions Regulator · 29 Oct 2026 · Accessed 29 Oct 2026
- Aon survey finds three-quarters of UK pension scheme trustees using AI Aon · 16 Oct 2026 · Accessed 29 Oct 2026