The AI policy void at major pension funds in 2026

If you sit in an investor-relations function at a public company in 2026, the question we keep getting is which institutional-investor expectations on AI you should be preparing the board to address. The honest answer is that the largest pension and sovereign-wealth funds have published surprisingly little on AI specifically, and that absence is itself the signal worth reading.

We expected, going into this piece, to find a corpus of AI position papers from the largest pools. We did not. Norges Bank Investment Management’s published position papers include statements on board diversity, related-party transactions, board independence, and CEO remuneration; the most recent paper is dated 2021, and there is no AI position paper. NBIM’s 2025 annual report mentions AI in the context of internal risk monitoring; that is operational disclosure, not portfolio expectation. The pattern repeats across the largest US public pension funds, the major Canadian pension plans, and the UK and EU pension landscape: AI is appearing in passing, in ESG-overlay frameworks, and in proxy-voting commentary, but not in the formal expectations document that institutional investors use to anchor portfolio-company conversations.

The structural read: AI is being rated by these investors, but the rating criteria have not been formally codified. IR teams preparing for institutional-investor conversations on AI are working against expectations the investors have not yet written down.

What the published corpus says, and where the silence sits

The largest sovereign-wealth and pension funds publish their stewardship priorities in formal expectations documents. NBIM publishes a position-paper series on specific governance topics. CalPERS publishes its sustainable-investments program priorities. The major Canadian plans (CPP Investments, OTPP, AIMCo) publish responsible-investment reports. The Dutch ABP / APG, the UK USS, and the major Nordic AP funds publish similar materials.

Across this corpus, the published AI content as of early 2026 is largely indirect. AI appears in three places:

• In broader technology-and-society overlays. Several funds have addressed AI as a theme inside a larger digital-rights or human-rights frame, often referencing the OECD AI Principles or the UN Guiding Principles on Business and Human Rights. The frame is portfolio-wide; the specific corporate expectation is not codified.
• In proxy-voting guidelines updates. Some 2024-2025 guideline updates added language about director responsibility for “emerging technology risk,” which extends the existing Caremark-style oversight expectation to AI without naming it. The expectation is implicit.
• In annual stewardship reports. Funds discussing engagement priorities for the year occasionally name AI alongside climate, executive remuneration, and cybersecurity. The discussion is forward-looking; the engagement record is thin.

What is largely absent: a formal, named AI position paper of the same shape as NBIM’s “Board diversity” paper or CalPERS’s climate engagement materials. The absence is consistent across the largest funds in 2026.

Why the absence is the signal

Three reasons explain the gap, none of them flattering.

The technology has moved faster than the formal-expectations process. Position papers at the largest funds typically take 12 to 24 months from internal initiation to publication, with extensive consultation. The 2022-2024 inflection in agentic AI capability has run faster than the publication cycle. Some funds have AI position papers in development; the publication date is not yet visible.

The expertise gap inside investment teams is real. Stewardship teams are heavily resourced on climate, governance, and human capital; AI expertise lives in operational risk teams or in external advisors. Institutional formal-expectations documents require domain expertise that has not yet been hired in at the same depth as climate.

The right altitude for an expectations document is not yet clear. Climate has the TCFD / IFRS S2 disclosure framework as its anchor; corporate expectations align to the framework. AI in 2026 has multiple competing anchors (NIST AI RMF, ISO/IEC 42001, EU AI Act conformity), none yet dominant. Funds reluctant to publish a position that ages quickly are deferring.

The combination produces a quiet 2026: large investors are forming opinions, raising AI questions in engagement meetings, voting on AI-adjacent shareholder proposals: large investors form opinions without yet having written down the standard the company is being rated against.

What this means for public-company IR teams

For an IR or Corporate Strategy lead preparing the board for 2026 institutional-investor engagement on AI, four practical implications.

Expect AI questions in engagement meetings without expecting a question pack. Major investor engagement teams are asking AI questions on a per-conversation basis, not from a published common-question framework. The questions vary; preparation against a single template is insufficient.

The Caremark-line oversight question will dominate. Even without a formal AI position paper, the director-supervision question (covered in our D&O insurance and AI supervision claim piece) is the most likely substantive area of investor probe. The board’s documented AI oversight system is the artefact investors will ask about.

EU AI Act readiness lands as an unspoken expectation. Funds with material EU exposure (Norwegian, Dutch, French, German) will read EU AI Act conformity as a baseline. US-domiciled funds with EU portfolio exposure will follow. Investors are unlikely to publish “we expect EU AI Act conformity” before the deadline; they are likely to ask about it in 2026 meetings.

Proxy-voting on AI-adjacent proposals is happening without published rationale. Several 2025 shareholder proposals on AI transparency and AI-related risk disclosure received material support; the funds supporting them did not always publish their reasoning. IR teams need to model the voting positions empirically rather than from explicit guidance.

The pension-fund engagement test

For a public-company board considering whether its current AI posture would survive institutional-investor engagement, three honest questions sit upstream of any published framework.

Could you describe the company’s material AI deployments to a stewardship lead in fifteen minutes, naming each deployment, its business criticality, the responsible executive, and the oversight mechanism? If the answer is “we’d need to pull that together,” the gap is the gap the investor will see.

Could you explain the company’s AI governance to an investor in the language of EU AI Act risk management and US Caremark oversight, even though the investor has not written either expectation down explicitly? Both vocabularies are now the substrate.

Could you connect the AI strategy presented to investors at the most recent Capital Markets Day to the AI risk register the audit committee reviews? The disconnect between IR-presentation AI and audit-committee AI is the most common preparation failure.

Three honest answers in the affirmative is the bar. Two affirmative answers is repairable with a quarter of focused work. One or zero is the territory where the engagement conversation goes badly.

What we are not claiming

We are not claiming that no major pension fund has published any AI material. The corpus is not zero (there are stewardship-report mentions, ESG-overlay references, and proxy-voting language updates). We are claiming that no published formal AI position paper exists from the largest funds at the same shape as their climate or governance papers, as of April 2026.

We are not claiming that the absence reflects investor indifference. The opposite is more likely: AI is being rated, but rated informally and against expectations not yet codified.

We are not predicting which fund will publish the first formal AI position paper. The candidates are several; the timing is uncertain.

What changes this read

Cadence on this piece is 60 days because position-paper publication cycles run on multi-quarter timescales. The three things that would change the verdict:

The first major fund to publish a formal AI position paper would shift the framing from “the absence is the signal” to “here is the explicit standard.” Watch for NBIM, CalPERS, the Dutch ABP, the Canadian CPP Investments, or the UK USS as the most likely first movers. A coordinated investor coalition publishing a multi-fund AI expectations document (analogous to Climate Action 100+ for climate) would change the engagement landscape immediately. EU AI Act enforcement landings in late 2026 would force funds with EU exposure to reference the regulation in their engagement priorities, even without a formal position paper.

We will re-test against NBIM position papers, CalPERS sustainable-investment publications, and PRI signatory updates on or before 30 Jun 2026.