US pension funds are sharpening their focus on director oversight of company AI usage, with three funds across the country adding language to their voting policies on holding them accountable for poor AI oversight.

San Francisco’s $33 billion employee retirement system was the first fund to add language to this effect. Trustees voted this year to approve amendments to its proxy voting policy stating that, “where there is evidence that insufficient oversight and/or management of AI technologies has resulted in material harm to shareholders”, the fund may vote against relevant directors.

San Francisco was joined by Vermont’s $7 billion pension investment commission in March, which amended its proxy voting policy to allow it to “review on a case-by-case basis the election of directors accountable for AI oversight after shareholders are deemed materially impacted by insufficient AI oversight or management”.

The issue is not just under consideration by funds in democrat-leaning states. Florida’s giant State Board of Administration (SBA) has adopted arguably the most progressive policy.

The San Francisco policy says the fund will not generally vote on directors over AI oversight or disclosures. By contrast, the $257 billion Florida fund said it may also oppose board members responsible for AI oversight “when there is evidence of failure to establish risk guidelines, review management proposals and establish board-level expertise”.

The Florida SBA also said it would back “reasonable” disclosure proposals on AI usage. A range of AI proposals have gone to the vote at corporate AGMs in recent years, although none have yet achieved majority support. Proposals asking for AI transparency at Netflix and Apple have come the closest, with 43.3 percent and 37.5 percent support.

Lindsey Stewart, director of stewardship research and policy at Morningstar Sustainalytics, told Responsible Investor that he did not expect such policies to become widespread.

“At a time when a lot of institutions are reducing the level of detail in their policies to allow themselves flexibility on voting actions, I’d still be surprised to see a lot of AI-specific policies on board director elections,” he said.

“But investors are still likely to continue engaging with companies to establish whether there’s appropriate whole-of-board competence in overseeing technology-related risks, whether that’s AI, cybersecurity or other matters.”

Read more @