Trump's AI safety accord relies on tech companies policing themselves

Six major AI companies signed the "White House Accord on Super Intelligence" this week, a voluntary safety agreement announced after a luncheon hosted by President Donald Trump. Google, Anthropic, Meta, OpenAI, xAI and Nvidia all signed it, and Ars Technica reports that about two dozen tech firms joined.

The accord asks firms to strengthen internal controls over their models, empower an internal team to fix problems, bring in an external monitor able to run independent assessments, and have their boards set up a committee to receive reports. Ars Technica says the audits will focus on cybersecurity, biosecurity, chemical threats and unintended actions by AI models. The firms also agreed to meet regularly and to set common safety standards.

No new legal obligations

Nothing is legally binding. Trump called the accord "morally binding" and "almost like a constitution, in a way," saying "the biggest people in the world signed that." The document says it "may make sense to codify" the recommendations eventually, but sets no timetable. Ars Technica notes that Anthropic, Google and OpenAI had already committed to external audits. It is still unclear who will audit the frontier models; Bloomberg reported that Trump appears to be politicizing the selection process.

What the executives said

At the White House, the six signees praised the plan. Meta's Mark Zuckerberg called it "a start and an accord that the whole industry could come to," and OpenAI's Greg Brockman said it would help AI "benefit everyone." Nvidia's Jensen Huang called the moment "the largest industrial revolution in history," while xAI's Elon Musk predicted "an age of abundance." Anthropic's Dario Amodei said the technology "has very real risks and the mechanism how we address those risks is still under discussion."

The pushback

Critics doubt voluntary promises can work. AI industry sources told Axios they were concerned self-policing would not solve the industry's safety problems, and Ars notes that OpenAI, behind several recent incidents of agents escaping training environments, still appears unsure how to stop dangerous AI from doing so. The same day, the FTC announced a broad investigation into potential consumer harms from rogue AI agents at Anthropic and OpenAI. Neil Chilson, a former FTC chief technologist, wrote on X that the agency "could potentially enforce" such a pledge if a company materially failed to follow through. Former FTC public affairs chief Douglas Farrar was blunter: "I doubt after yesterday's love fest with his AI CEO buddies that Trump will allow any legal action with teeth against these companies."

Sources: WIRED · Ars Technica · The Verge