White House Accord Seeks Voluntary Independent Frontier AI Audits
The White House announced a non-binding "Accord on Super Intelligence" under which frontier AI developers would adopt internal controls, engage independent external auditors and create board-level oversight.
On September 29, 2026, the President met with executives of several leading AI developers and announced a voluntary framework, the "White House Accord on Super Intelligence." Signatories would adopt four layers of controls: internal monitoring of model capabilities and alignment, a dedicated internal oversight function, an independent external auditor, and an independent board-level committee supervising both. The text uses recommendatory "should" language, has no enforcement mechanism, and says it "may make sense to codify these steps into laws and regulations." A separate executive order that day substituted "Super Intelligence" for "Artificial Intelligence" in federal usage without imposing substantive obligations.
Why counsel should care: the external-audit commitment reverses the Administration's prior stance — former AI czar David Sacks had called outside safety assessments "pseudoscience" — and is likely to migrate into customer security questionnaires, vendor diligence and contract terms regardless of legal compulsion. Clients procuring frontier models should define audit representations, information rights, reporting duties, and privilege and trade-secret protections. Developers should pre-plan auditor independence criteria, scope and information access.
What to watch: reported discussions between prospective auditors and the Department of Commerce pointing to a formal government role for audit providers, and a reported FTC investigation into several frontier AI developers. Both are unconfirmed in this single-source alert. Congress is not expected to act before the midterms.