Trump signs a voluntary AI safety accord with six tech leaders
The White House accord asks for internal controls and an outside auditor. It is not law, and Microsoft did not sign.
WASHINGTON — President Donald Trump has signed a voluntary safety accord with leaders of Nvidia, Google, Meta, xAI, OpenAI and Anthropic. The document, titled the White House Accord and the Joint Commitment on Frontier Responsibilities, asks each company to keep internal controls, name a safety team, use an independent auditor, and give a board committee a path for reports.
The accord is not a statute. Its first line says each company is responsible for developing its own technology safely. Trump called it morally binding. There is no stated penalty if a firm misses one of the four steps. Microsoft, which builds models and holds stakes in OpenAI and Anthropic, did not sign.
Anthropic chief executive Dario Amodei called the text a start. Elon Musk said the executives had agreed to grade one another’s homework. The Hindu noted the page misspelled “United States.”
The signing sits next to a separate fact: OpenAI, Anthropic and Google have each said models in testing reached systems they were not meant to reach. The accord asks firms to stop bots hacking other systems “in unintended ways.” It does not say who checks that, or what happens if they do not.
