President Donald Trump released a one-page artificial intelligence accord Tuesday, signed by the leaders of the biggest AI companies, that asks them to voluntarily police their own systems as fears mount over the technology's risks.
Trump posted the document to his Truth Social platform after gathering executives at the White House, describing it as "morally binding."
It was signed by Trump and the heads of Google, Anthropic, Meta, OpenAI, X and Nvidia, according to the document. OpenAI was represented by President Greg Brockman, alongside Google's Sundar Pichai, Meta's Mark Zuckerberg, Anthropic's Dario Amodei, Nvidia's Jensen Huang, and Elon Musk, whose xAI is now part of SpaceX.
The accord lays out four voluntary measures. It asks companies to maintain robust internal controls to monitor their models, particularly around cybersecurity, biosecurity and chemical threats, and to ensure the systems "do not hack or access technical systems in unintended ways."
It also calls for an internal team to verify those controls are working, an external auditor to independently assess them, and an independent board committee to oversee the whole arrangement. The document says the steps could eventually be codified into law.
Sep 23 Sep 25 Sep 27 Sep 29 Sep 30 $85.3k $84.4k $83.5k $82.7k 24h High High $85,518 24h Low Low $82,951 Vol Vol $1.5B Market projections Odds by Myriad Today Below $84,000 Below $84k 53 % chance This week Below $84,000 Below $84k 51 % chance This month Below $84,000 Below $84k 53 % chance → Buy Bitcoin with USDT Powered by Jupiter $ 50 $ 100 $ 500 Buy Price data by CoinGecko CoinGecko More Bitcoin news and projections → The pact carries no penalties for falling short, and critics were quick to note its nonbinding nature and thin detail. Trump, who has previously downplayed AI risks, said he would set up a roughly 10-person committee to oversee the effort and name an official to lead it.
The accord lands after a bruising stretch of incidents in which AI agents slipped their leashes.
Such agents have repeatedly escaped their creators' control , an OpenAI agent breached an Australian government website , and Australia's parliament has since summoned OpenAI’s Altman and Anthropic’s Amodei to testify. The timing is pointed: by several accounts, the accord would not have required OpenAI to disclose the Australia breach any sooner than it did.
The self-policing approach reflects the administration's broader skepticism of hard AI regulation, even as figures like Anthropic CEO Amodei push for the industry to slow down. It also arrives alongside debate over harder safeguards, including proposals for an AI "kill switch" that Nvidia has weighed in on.