Only 39% of Fortune 100 boards have any form of AI oversight structure. Only 13% of S&P 500 companies have a single director with documented AI expertise. With BlackRock, Allianz, and Glass Lewis updating 2026 stewardship guidelines to require demonstrated AI literacy — and the SEC formally recommending AI disclosure standards — the proxy season is no longer a future concern. It is a present accountability test. This checklist translates the governance gap into governance action. Execute it before your next board meeting.
1. Establish Structural AI Oversight
- Assign AI oversight responsibility explicitly to a named committee — Audit, Risk, or Technology — with AI documented in that committee's charter.
- If no committee charter currently references AI, amend it at the next available board meeting.
- Consider whether a dedicated Technology or AI Committee is warranted given your organization's level of AI deployment.
- Appoint a committee chair who will own the AI risk reporting relationship with management.
2. Map Your Organization's AI Exposure
- Request an inventory of every AI system that influences strategy, financial reporting, compliance, talent decisions, cybersecurity, and customer-facing operations.
- Classify each system by materiality: low-risk productivity tools versus high-impact systems that alter decisions or create legal exposure.
- Identify which AI systems, if they failed or produced biased outputs, would generate regulatory, reputational, or Caremark-level liability.
- Confirm whether any third-party or vendor AI is embedded in material business processes.
3. Demand a Structured AI Risk Inventory from Management
- Reliability: How are AI systems tested before deployment? What is the failure rate and escalation protocol?
- Fairness: What data sets are used for training? How are models tested for bias? Who conducts independent review?
- Security: What is the attack surface of AI-connected systems? How are adversarial inputs managed?
- Compliance: Are third-party AI vendors contractually accountable to regulatory standards? Who tracks evolving AI regulation?
4. Set Transparency Standards for the Boardroom
- Establish a board protocol stating that AI outputs are inputs, not conclusions — human judgment closes every material decision.
- Require management to disclose when AI shaped strategic recommendations, earnings narratives, workforce analyses, or customer risk assessments presented to the board.
- Flag any AI-generated analysis that lacks disclosure of model limitations, training data sources, or confidence intervals.
5. Close the Director Literacy Gap
- Assess each director's AI literacy against the oversight responsibilities the board now carries.
- Enroll directors in structured AI governance programs — NACD's AI oversight certification or equivalent — and document completion in proxy disclosures.
- When evaluating new director candidates, treat AI literacy as a priority skill alongside financial and sector expertise.
- Schedule at least one dedicated board education session per year focused exclusively on AI risk, not AI opportunity.
6. Prepare for Proxy Season Disclosure
- Confirm your proxy statement addresses: how AI is defined within your organization, which board body oversees AI deployment, and the material effects of AI on internal operations and customer-facing activities.
- Review against the SEC Investor Advisory Committee's December 2025 AI disclosure recommendations.
- Benchmark your disclosure against Glass Lewis and ISS 2026 stewardship guidelines before the proxy is filed.
- Confirm that documented director training and oversight frameworks are included — proxy advisors now treat this as a baseline expectation.
Advisory Note: The Caremark doctrine extends to AI. A board that fails to implement a reasonable AI reporting and monitoring system — or that receives red flags and does not act — faces shareholder derivative exposure. Documented evidence of the actions above is not a formality; it is your legal defense. Begin building that record now, not after an incident forces the conversation.