Trump renamed AI. The safety deal signed beside it matters more.

President Trump signed an executive order on September 29 telling every federal agency to stop saying "artificial intelligence." From now on, official letters, websites, reports and policy documents have to call it "Super Intelligence," or SI. The White House says the new name "conveys the true capabilities" of today's technology.

The order covers non-statutory documents only, so the many laws that already say "artificial intelligence" don't change. It also asks the President's science and technology adviser to propose a federal definition of Super Intelligence and suggest further steps. Until that definition arrives, the government is using a term researchers usually save for AI that beats people at almost everything, and applying it to every chatbot and spam filter it buys.

The rename got the headlines. The document signed at the same White House summit says more about how AI will be governed.

Six companies put their names to a one-page Accord on Super Intelligence: Google, Anthropic, Meta, OpenAI, Nvidia and xAI, represented by Sundar Pichai, Dario Amodei, Mark Zuckerberg, Greg Brockman, Jensen Huang and Elon Musk. It asks every company training frontier models to run four layers of oversight. First, internal controls that track a model's capabilities and alignment, including cyber, biological and chemical risks. Second, an internal team that checks those controls work. Third, an independent outside auditor. Fourth, a board committee that receives the reports and makes sure problems get fixed.

That sounds serious until you look at what's missing. The accord has no penalties for breaking it. Companies don't have to report safety incidents to the government. They pick their own auditors, and nothing requires them to publish what those auditors find. There's no compliance date either. The text only says that "over time" it may make sense to turn these steps into law.

Trump called the agreement "morally binding." Senator Richard Blumenthal had a different description: “completely voluntary and totally secret.” He argued it gives Congress a free pass to keep not legislating.

Meta signed even though it is not part of CAISI’s pre-deployment testing program. Signing this accord does not itself require participation in that program.

The timing makes the voluntary approach harder to defend. A day later, The New York Times reported that the FTC had opened an investigation into OpenAI and Anthropic, two of the signers, over possible harm to consumers, including from AI agents that went beyond what they were told to do. One part of the government is asking companies to police themselves while another part is checking whether they did.

Two things come next. The White House still has to say what "Super Intelligence" actually means, and the FTC is expected to send formal demands for documents to OpenAI and Anthropic in the coming weeks. The accord will matter only if it changes what those documents show.