Six AI firms sign voluntary White House safety accord
Leaders of Google, Anthropic, Meta, OpenAI, Nvidia and SpaceX signed an accord with Trump calling for outside audits of frontier models. It carries no penalties.
What it asks
Every company training frontier models is asked to run four layers of control: internal controls, an internal team testing them, an independent external auditor and an independent board committee. Covered risks are cyber, biological and chemical threats.
Its limits
The one-page text is voluntary, with no penalties, deadlines or duty to publish audit results, and companies pick their own auditors. Trump called it "morally binding".
More articles
News · Policy
