The most dangerous governance failure in 2026 is not the failure to innovate. It is the failure to verify. Boards that have spent the last three years building AI governance declarations (responsible AI policies, ethics pledges, AI principles statements) and treated those declarations as governance have built precisely the architecture the SEC’s enforcement record now identifies as inadequate.
The distinction matters. A declaration is a statement of intent. Verification is evidence that the intent has been executed. The board that declares “our AI systems are accurate and ethical” and the board that can demonstrate, with current technical evidence, that specific investor-facing AI claims reflect the actual performance of specific systems are not in the same governance posture. The first board has a policy. The second has a governance architecture.
This is the Declarative Board Failure Pattern: the organization completes the declaration and treats it as the governance. The SEC’s enforcement record of 2025 makes clear that the market, the courts, and the regulators are now treating the gap between declaration and verification as material, and the consequences for organizations that have not closed it are specific, legal, and quantifiable.
The SEC’s Cyber and Emerging Technologies Unit (CETU), established in February 2025, pursued two landmark AI washing cases in its first operational year. Presto Automation’s investor communications described its voice AI system as reducing or eliminating the need for human labor in restaurant drive-through operations. The actual operational record showed that human agents completed the vast majority of transactions the system initiated. The SEC’s settlement in January 2025 did not turn on the question of whether Presto had a responsible AI policy. It turned on the gap between the investor-facing claim and the technical evidence.
The Nate Inc. case, filed April 9, 2025, went further. Albert Saniger, the founder and former CEO, faced parallel SEC civil charges and DOJ criminal prosecution. The organization had raised $42 million from investors based on AI capability claims that internal records demonstrated were inaccurate. The criminal prosecution converted a governance failure into a personal criminal exposure.
These are not regulatory edge cases. They are the beginning of a docket. CETU is a standing enforcement unit. Its first-year enforcement pattern is a template, not an exception.
The September 2025 Delaware Court of Chancery ruling in Giuliano v. Grenfell-Gardner (the Teligent case, C.A. No. 2021-0452-KSJM) reaffirmed the Caremark obligation with new urgency: where a board fails to build information and reporting systems for the regulatory obligations most central to its business model, derivative liability is clearly stated. The court found that Teligent’s board (of a generic pharmaceutical manufacturer) had built no system to track FDA decisions that could halt production, despite FDA compliance being the company’s most existential external constraint. The principle applies to AI governance without modification: the board of an organization that makes AI capability claims to investors must build oversight architecture for those claims or face the same documentation failure. Board-level oversight of AI claims is a fiduciary obligation, not a governance aspiration. A board that receives management’s AI governance presentations without adversarial questions, without documented evidence requests, without a defined oversight framework is not exercising oversight. It is receiving information.
The distinction between receiving information and exercising oversight is where most board governance fails. The Boeing derivative litigation established this principle a decade ago in the context of airplane safety systems: board members who received safety reports without adversarially interrogating them did not satisfy Caremark. The Teligent ruling of September 2025 reaffirmed the same logic, making clear that the Caremark oversight duty attaches to whatever regulatory obligations are most central to the organization’s business model, and that the obligation is not satisfied by passive receipt of management reports.
This creates a specific governance obligation that declarations cannot satisfy: the board must be able to demonstrate, from its own records, that it asked the questions that would have surfaced a governance gap before enforcement arrived. The board that receives an AI governance update and responds with approval has a documentation problem. The board that receives an AI governance update and responds with documented adversarial questions (did we verify this claim? What is the technical evidence? Who is accountable for the verification? What happens if a gap is found?) has produced the Caremark documentation.
The argument from innovation is still present in boardrooms, and it deserves a direct response.
Some boards resist verification architecture because they believe it will slow down their organizations’ AI adoption. The concern is legitimate on its surface: if every AI capability claim requires technical verification before publication, some claims will be slowed or stopped. This is true. It is also the correct outcome. A claim that cannot be verified against current technical evidence should not be published.
The Expectation Elevation Model clarifies what is actually at stake. The compliance-minimum posture builds verification architecture sufficient to avoid enforcement exposure. The elevated posture builds verification architecture that becomes a competitive asset: an organization whose AI capability claims are accompanied by independent technical verification can make those claims with a confidence that organizations without verification infrastructure cannot match. The Barrios et al. research from NBER Working Paper 33645 (April 2025) quantifies this directly: the conflict-of-interest discount in the marketplace of ideas means that claims with visible independent substantiation carry more market weight than claims without it.
The verification imperative is not an obstacle to innovation. It is the architecture that makes innovation claims worth making.
The specific governance posture that the Caremark expansion requires of boards is not complicated, but it does require deliberate design. The board must designate a named accountable individual for AI claims governance, with documented authority, not advisory authority. The board must receive AI governance reporting on a defined cadence, and the reporting must include the verification status of specific investor-facing claims, not the general posture of the organization’s AI ethics program. The board must ask adversarial questions at each reporting cycle and document that it did.
Most boards are not doing this. The National Association of Corporate Directors’ 2026 Governance Outlook found that while 62% of directors now include AI as a board agenda topic, the depth of engagement varies substantially. Having AI governance on the agenda is not the same as exercising AI governance oversight. The agenda item is the declaration. The adversarial interrogation is the governance.
The leaders who will define the governance standard of the next decade are not the ones who built the most comprehensive AI ethics declarations. They are the ones who built the verification systems that made those declarations true and demonstrable, before enforcement required it, before a PE buyer asked for the documentation, before a derivative plaintiff found the evidentiary gap. Not built in response to an enforcement action. Built from conviction that the gap between declaration and evidence was a problem worth solving before anyone else named it.
The Legacy Test for a board is not whether the declarations were compelling. It is whether the successors they leave behind inherit a verification architecture that continues to protect the organization after the current board is gone, a system that operates because it was embedded in governance structure, not because the people who built it are still in the room.
This analysis was developed in the Ethics as an Advantage: Why Trust Will Be the Most Valuable Currency in the 2026 Economy Executive Leadership Playbook, which includes functional white papers for each C-suite role.
Touch Stone Publishers Limited | 2026 | touchstonepublishers.com