The Pensions Regulator (TPR) has set out expectations for the responsible use of artificial intelligence (AI) in workplace pensions, signalling that trustees remain accountable for outcomes even as pension funds and service providers increasingly deploy AI tools across administration and member services.
In updated guidance and its AI plan, TPR emphasises that AI should be used to support, rather than replace, human decision-making in pension scheme governance.
The regulator makes clear that responsibility for compliance, operational resilience and member outcomes cannot be delegated to technology providers, even where systems are automated or externally managed.
The intervention comes as pension schemes, administrators and asset managers explore the use of AI across a widening set of functions, including member communications, fraud detection, data processing and operational efficiency improvements.
TPR’s position effectively brings AI into the mainstream of trustee oversight expectations, placing it alongside other core governance concerns such as cyber risk, data quality and third-party risk management.
Nausicaa Delfas, chief executive officer of TPR, said: “AI has the potential to transform pensions for the better: improving how schemes are run, how members are supported and how the system as a whole delivers value.
“But trust is the most valuable asset in our system, and that trust depends on the safe and responsible adoption of AI in members’ interests.”
While the regulator stops short of introducing new binding rules, the tone of the guidance signals a shift towards formalised scrutiny. Trustees are expected to understand how AI systems are being used within their governance chain, particularly where third-party administrators or service providers are deploying generative or predictive tools on their behalf.
The practical implication is that schemes may need to revisit oversight frameworks to ensure AI usage is documented, explainable and subject to appropriate human controls. This includes clarity over how data is used, how outputs are validated and how errors or bias in automated processes are identified and corrected.
Delfas said: ”Our message to trustees, administrators and scheme managers is clear: act now. Put strong governance in place, invest in data quality, understand where and how AI is being used in your scheme, and protect your members from AI-driven fraud.”
TPR’s guidance also underscores a broader regulatory trend towards principles-based oversight of emerging technologies, rather than prescriptive rules.
By framing AI as an extension of existing governance duties, the regulator is effectively signalling that trustees do not need to become technology experts, but they do need to ensure appropriate controls are in place and that ultimate accountability remains clearly defined.
As AI adoption in pensions accelerates, attention is likely to shift towards how regulators test compliance in practice. For trustees, the immediate challenge is less about whether to adopt AI, and more about how to demonstrate that its use is properly governed, transparent and aligned with fiduciary responsibilities.
