White House AI agreement missing provisions on independent auditor qualifications and disclosure of results

The White House’s superintelligence accord, signed on September 29, 2026 by President Donald Trump and the heads of Google, Anthropic, Meta ($META), OpenAI, xAI and Nvidia ($NVDA), commits advanced-AI developers to internal safety controls, internal review and cooperation with independent external evaluators. It leaves undefined, however, the criteria for judging evaluators’ qualifications and independence, and how their findings must be disclosed. The four-stage safety framework therefore lacks rules on who may audit frontier models and what the public learns from those audits. The gap matters because the accord’s credibility rests on outside scrutiny of the same companies building the most capable models. What to watch next is whether the signatories or the White House issue detailed auditor standards and disclosure requirements.