Board oversight of artificial intelligence has crossed from best practice to fiduciary expectation. The National Association of Corporate Directors now frames AI oversight as a core director duty, proxy advisors weigh it in 2026 vote decisions, and the SEC keeps sending comment letters. Yet fewer than three percent of S&P 500 directors disclose any AI expertise. The standard has moved. Most boards have not.
PRIORITY 9 | SILO: MARKET SENTIMENT
The institutions that grade boards now treat AI oversight as a fiduciary duty, while director expertise and committee charters lag far behind the risk disclosures already on file.
The Signal
The National Association of Corporate Directors has published Director Essentials: Implementing AI Governance, and its framing is the news. AI oversight is no longer presented as a technology option for the audit committee to consider. It is presented as a core fiduciary responsibility spanning AI strategy, capital allocation, risk, and board competency.
This closes a loop that has been forming for a year. When a director institution, the proxy advisors, and the SEC all define the same duty in the same season, the duty exists in practice, whether or not a court has named it yet. The question for a board is no longer whether AI oversight is expected. It is whether the board can show it.
The Evidence
The disclosure has raced ahead of the capability. The Conference Board reports that the share of S&P 500 companies disclosing AI as a risk rose from 12 percent to 83 percent between 2023 and 2025. Over the same window, the share of directors disclosing AI expertise moved from 1.5 percent to 2.7 percent. Boards have learned to name the risk. They have not yet staffed the oversight of it.
The structural picture matches. NACD data shows roughly one in four boards has written AI oversight into a committee charter, and an ISS-Corporate analysis found that 54 percent of the S&P 100 disclosed board-level AI oversight in 2025 proxies, but only 28 percent disclosed both oversight and a formal policy behind it.
The pressure is not theoretical. The SEC has issued 92 AI-related comment letters to 56 companies since 2021 and received a formal petition for mandatory AI governance disclosure rules in February 2026. Glass Lewis has named AI oversight the defining governance theme of the 2026 proxy season. The graders are aligned. The gap between what companies disclose as a risk and what boards can prove they govern is now the exposed surface.
The Strategic Implication
Defensive Risk. The exposure sits with audit committee chairs and lead independent directors at companies that have already filed AI as a material risk factor while disclosing no board structure that oversees it. That mismatch, a named risk with no oversight record behind it, is the exact fact pattern a Caremark plaintiff pleads as a red flag consciously ignored, and it is a ready vote-against vector under the 2026 proxy-advisor standard. The window is the next 10-K and proxy cycle. The defensive move is to write AI oversight into a named committee charter now and document the information flow that reaches the board, rather than letting a proxy advisor or a plaintiff define the standard first.
Offensive Advantage. The same season that exposes the unprepared rewards the prepared. A board that seats real AI competency, charters a committee to own the risk, and shows a working reporting line can state in its proxy what most peers cannot: the oversight exists and it functions. In a year when 85 percent of boards cannot say that credibly, the ones who can convert a compliance scramble into an investor-trust signal that competitors cannot quickly copy.
There is a boundary worth naming. A board that adds AI oversight only because Glass Lewis would otherwise recommend against its directors has not adopted a standard. It has complied with one. The Governance Boundary Principle holds that the board must own the oversight standard itself, not receive it. Ownership is what still stands the season after the proxy advisor moves on to the next priority.
The distance between an AI risk factor and a board that can actually govern the risk is the liability. Close it before someone else measures it for you.
Board chairs and audit chairs: Take the Board Fiduciary AI Stress Test at touchstonepublishers.com/board-fiduciary-assessment