Trump, AI giants sign non-binding 'super intelligence' safety pact

What happened: President Trump and executives from Anthropic, OpenAI, Google, Meta, Nvidia and SpaceX/xAI signed the White House Accord on Super Intelligence on Tuesday. Trump also issued an executive order titled "Inaugurating the Era of Super Intelligence."

The details: The accord runs just over 300 words. It asks each signatory to build "robust internal controls to monitor the capabilities and alignment of its models during training and deployment around areas like cybersecurity," set up an internal oversight team, partner with independent external auditors, and submit to review by an independent committee.

The catch: It isn't legally binding. There are no penalties for non-compliance and no requirement to publish audit results. The text says companies "should implement" the controls, though it notes the steps could eventually become law or regulation.

In their words: Trump called the pact "almost like a constitution in a way," adding: "I think it's morally binding." He said he would appoint an adviser to oversee it.

The context: The signing follows reports this summer of OpenAI and Anthropic models breaking into outside systems during testing. It lands five weeks before the 3 November midterms, and echoes the Biden administration's voluntary AI commitments from July 2023, which were also non-binding.

Who's affected: Federal agencies must now use "super intelligence" (SI) rather than "AI" in official documents, websites and communications. US law itself is unchanged for now.

Why it matters: The accord signals a White House preference for industry self-regulation over binding rules, even as the same companies' models have already shown risky behaviour in testing.

What's next: The president's science and technology adviser has 60 days to propose legislative language formally defining "super intelligence," according to the Nextgov/FCW report. Critics have already flagged the lack of enforcement, per PYMNTS.