Executive Summary
Glass Lewis can now recommend a vote against directors when a material AI failure meets a board that disclosed no AI oversight, and that policy is live for the 2026 proxy season already underway. Only 15 percent of the S&P 500 disclose board-level oversight of AI. The exposure is concentrated and specific, and it becomes acute the moment an AI incident surfaces inside a company’s annual meeting window.
The Signal at a Glance
PRIORITY 9 | SILO: Institutional
A live 2026 proxy-advisor policy converts the board AI-oversight gap into a director-election risk, against an S&P 500 where 85 percent disclose no such oversight.
The Deep Dive
The Signal
Glass Lewis applied its AI-oversight policy to shareholder meetings held after January 1, 2026. The policy is threshold-based: absent a material AI-related incident, Glass Lewis will not recommend against directors on AI grounds. When shareholders are materially harmed by insufficient AI oversight or management, it may recommend against the responsible directors. The structure mirrors the cybersecurity-oversight policy Glass Lewis introduced in its 2024 guidelines, which is the template for how a disclosure expectation hardens into a vote.
The reason this is a 2026 signal, not a 2027 one, is timing. The policy is operative now, in the current proxy season, while the disclosure base that would protect a board is thin. The trigger is not a missing AI policy. The trigger is a material incident landing on a board that cannot point to any disclosed oversight structure when the proxy advisor and its institutional subscribers look.
The Evidence
The disclosure gap is documented by Glass Lewis itself. In a March 2026 analysis of S&P 100 proxy statements, Glass Lewis found that 54 percent disclosed board-level AI oversight and only 28 percent disclosed both oversight and a formal AI policy. Across the broader S&P 500, the SEC Investor Advisory Committee found that just 15 percent disclose information about board oversight of AI.
That second figure comes from the recommendation the SEC Investor Advisory Committee approved on December 4, 2025, by a vote of 14 in favor, 2 against, and 2 abstaining. The recommendation asks the Commission to require issuers to define AI, disclose board oversight mechanisms, and report material AI effects on operations and on consumer-facing matters. SEC Chair Paul Atkins and Commissioner Hester Peirce both signaled they do not see a need for Commission guidance, which means the federal disclosure mandate is unlikely in the near term. The proxy advisor, not the regulator, is the active enforcement vector in 2026.
The incident base that supplies the material-harm trigger is already accumulating. Glass Lewis cites a December 2025 warning to large technology companies from a coalition of US attorneys general over AI outputs, and a January 2026 UK regulatory investigation into X over Grok. Nine of 29 technology-related shareholder proposals in the 2025 season dealt explicitly with corporate AI use.
The Strategic Implication
This is the Governance Boundary Principle applied to AI. The board’s job is to own oversight and prove it owns oversight, not to manage the models. A disclosed committee charter is the cleanest evidence that the board has taken the oversight side of that boundary and held it. The proxy advisor is, in effect, now policing the boundary: it asks not whether the board ran the AI well, but whether the board can show it governed the people who did.
Defensive Risk. The director most exposed is the chair of the committee a board would name as its AI-oversight owner: the audit or technology committee chair at any S&P 500 company that has not yet disclosed an AI-oversight structure. The mechanism is specific. A material AI incident becomes public. An institutional holder routes its vote through Glass Lewis. The recommendation lands against the named director, who has no disclosed oversight record to cite in defense. The window is the company’s own annual meeting, anytime in the 2026 proxy calendar, because the policy is already in force. The responsible move is to seat AI oversight in a named committee charter and disclose the structure in the next proxy statement, before an incident forces the question.
Offensive Advantage. The 15 percent of the S&P 500 that already disclose board AI oversight hold a recommendation-proof position the other 85 percent cannot quickly assemble under incident pressure. A board that documents its oversight structure now converts a latent liability into a visible governance asset, the kind that survives an activist’s slate math and a proxy advisor’s case-by-case review. Lockheed Martin’s 2025 proxy, which split AI oversight across three named committees and identified directors with AI expertise in its skills matrix, drew no AI-related shareholder proposals in a year when peers did. Disclosed oversight is not a compliance cost. It is the cheapest activist deterrent available before the incident, and the only one unavailable after it.
A disclosed charter protects only the board that actually governs behind it. The document is evidence, not the oversight itself, and the boards that survive the incident as well as the recommendation will be the ones whose directors were in real contact with the people running these systems long before a proxy advisor came asking. The disclosure is the floor. Acting before the incident, rather than after the recommendation, is the difference.
Touch Stone Publishers