President Donald Trump and leaders of major AI companies signed a voluntary accord on September 29 covering internal reviews, independent outside evaluations and board oversight. The companies also commit to curbing unintended technical access by AI systems. The announced arrangement creates company commitments, while the participants have not announced a binding statute or completed government audit.
Artificial Intelligence··Morning
Companies commit to internal and outside reviews
US President Donald Trump announced a voluntary AI safety accord after meeting company leaders at the White House on September 29. Executives from OpenAI, Anthropic, Meta, Google and NVIDIA attended. Each company committed to establishing strong internal controls and having an independent outside evaluator examine whether they work. Findings are also meant to reach company boards. These are commitments by the signatories under the announced voluntary arrangement, rather than requirements imposed by a new law.[1], [2]
Boards are assigned a review role
The accord also calls for a committee on each company board to examine findings from internal and outside reviewers. It addresses the risk that AI tools could hack or access technical systems in unintended ways. That provision concerns actions a system might take through tools, beyond the accuracy of a model answer. The announcement describes the review structure, but does not establish a completed public audit, a detailed common test method or a penalty for a company that falls short of its commitment.[1]
Possible legislation remains separate from the accord
The signatories said these steps could eventually be written into laws or regulations. At the press conference, Trump said he was considering a ten-person body to watch over AI safety, without naming its members or powers. The White House meeting followed renewed debate about oversight of advanced models and reports of agents taking unauthorized technical actions. What was announced is a common commitment to layers of company review. Any government oversight body and any binding legal rule would require separate decisions.[1]
