A portfolio can hold investments across several sectors and still depend heavily on the same economic story.
The UK’s Pensions Regulator is asking trustees to examine that possibility as AI-related investment expands.
Guidance published October 22 calls for assessment of direct and indirect AI exposure across entire portfolios. Trustees are asked to use severe-but-plausible scenarios, reconsider whether diversification would hold if correlations changed, and challenge advisers and managers about the implications. The Pensions Regulator: October 22 guidance announcement.
Portfolio-wide exposure and diversification
The guidance concerns investment governance. It does not direct every scheme to sell AI-related assets or establish that a particular market outcome will occur.
Its central point is that familiar asset labels may conceal shared dependencies.
For example, an illustrative portfolio could hold technology companies, infrastructure investments and lenders whose prospects all depend partly on continued AI expansion. The holdings look different, but a common shock could affect them together.
A decision record that tests assumptions
For pension and fiduciary counsel, the useful question is how trustees demonstrate that they have examined those connections.
A sound decision record should identify the information considered, the assumptions tested and the reasons for accepting or changing the portfolio’s risk profile. Asking an adviser for reassurance is less informative than asking what happens under a defined adverse scenario.
The regulator also calls for proportionate action reflecting each scheme’s circumstances. The Pensions Regulator: October 22 guidance announcement.
Proportionate oversight
That requires more than adopting a standard AI-risk paragraph. The analysis should relate to the scheme’s actual investments, objectives and capacity to absorb losses.
Canadian trustees must assess their own governing obligations. The UK guidance nevertheless provides a timely comparison for examining risks that cross conventional sector boundaries.
The practical objective is informed oversight, rather than certainty about the future of AI.
Diversification needs to be tested against shared risks, not merely counted by the number of holdings.
Related coverage: Banking Regulators Ask Whether Autonomous AI Agents Can Be Controlled.
Sources: The Pensions Regulator: October 22 guidance announcement .