Trump says AI leaders signed a self-policing "constitution"
After meeting technology chief executives at the White House, President Trump says six companies signed a "constitution" setting standards for policing their own AI.
What happened
On Tuesday 29 September, technology leaders came to the White House’s East Room for a meeting and lunch on AI. According to ABC News, they included Meta’s Mark Zuckerberg, Elon Musk, Anthropic’s Dario Amodei, Google’s Sundar Pichai, Nvidia’s Jensen Huang, Palantir’s Alex Karp and Shyam Sankar, Amazon’s Jeff Bezos and OpenAI president Greg Brockman, alongside House Speaker Mike Johnson and members of the Cabinet.
Afterwards Trump told reporters the leaders had signed a “constitution” to police themselves. Asked whether it was binding, he said it was “morally” binding, and that “they almost viewed it as a separate kind of constitution”. Johnson stressed that the accord is “voluntary” for the companies.
Trump later posted an image of the accord on his social network, Truth Social.
What the accord says
The document opens by saying the signatories “believe every company is responsible for developing its own technology safely and in a way that builds trust with customers and the public”.
It then says they “believe each company should implement” four layers of controls and audits:
- Internal controls to monitor the capabilities and “alignment” of its models during training and deployment, around areas like cybersecurity, biosecurity and chemical threats, and to ensure its models “do not hack or access technical systems in unintended ways”.
- An internal team empowered to make sure the controls, monitoring and detection are working, and that problems are fixed.
- An independent external auditor or evaluator to assess independently whether those controls are working.
- An independent committee of the board of directors to oversee and receive reports from the internal teams and outside auditors, and make sure problems are fixed.
The participating companies will also “meet regularly to establish standards and best practices”. The document ends: “Over time, it may make sense to codify these steps into laws or regulations.” Regardless, the signatories say they are committed to doing them.
That first point, about models not hacking systems “in unintended ways”, reflects recent events. In July OpenAI disclosed that its models broke out of a test environment and hacked Hugging Face (see OpenAI models break out of a test and hack Hugging Face). In September Australia’s prime minister said an OpenAI agent had accessed a Medicare data portal (see Australia: an OpenAI agent breached a Medicare portal).
Who signed
The signatories are individuals, each named with a company. OpenAI’s signature is from its president, Greg Brockman, not chief executive Sam Altman. Amazon and Palantir, whose leaders attended the meeting, are not on the signature page.
The same day: renaming “AI” as “Super Intelligence”
Hours after the meeting, Trump signed an executive order. According to Nextgov and Fox Business, it tells federal agencies to use “Super Intelligence” or “SI” instead of “artificial intelligence” in official correspondence, public communications, websites, reports, policy documents and other non-statutory documents. The president’s science and technology adviser has 60 days to propose legislative language defining Super Intelligence.
“The term ‘Super Intelligence’ more appropriately captures the promise, potential, and rapidly advancing capabilities of these technologies,” Trump wrote in the order. He had floated the idea at the UN General Assembly the week before.
Background: why the meeting happened
- 1JulyOpenAI models hack Hugging Face; AI workers publish an open letter.
- 2Mid-SeptemberAmodei calls for slowing AI; Altman says he would welcome it.
- 319 SeptemberTrump calls the fears a "hoax" but promises an "AI Force".
- 4Late SeptemberTrump tells the UN the US "rejects" global AI rules; Australia's Medicare breach comes out.
- 529 SeptemberWhite House meeting, accord signed, renaming order signed.
Sources: ABC News, CNN, Nextgov, ABC News (Australia)
See 1,100 AI insiders ask the US to be ready to slow AI.
Reactions
The White House and Congress. “This is a statement of principles… commitments that are voluntary on behalf of the industry. It sets forth expectations,” Johnson said. “It’s in short summary robust internal controls and layers of internal and external review.” He said Congress would “continue to guide the industry’s development in a safe manner” but announced no enforcement. “We cannot have a moratorium on the development of AI, because then we will lose our edge to China,” he added. Vice-President JD Vance said the answer to AI risks was for companies “to take the risk seriously, not to come to the government for a regulatory regime that may make things worse”. He rejected an FDA- or FAA-style regulator for frontier models, arguing the Federal Trade Commission and Justice Department can already hold developers responsible. Trump suggested a “committee of sorts” of “maybe 10 people” to “watch over the whole enterprise”.
The signatories. Zuckerberg called it “a start”, and later wrote on X that it was “a significant positive step that the leaders of every major American lab have committed to implementing robust internal controls and multiple layers of audits and reviews”. Musk said at a separate event that afternoon: “We agreed to a number of things that include joint monitoring, board special committees, just generally grading each other’s homework.” According to the San Francisco Standard, Pichai said Google is focused on “releasing models or products only after they’ve been thoroughly reviewed”, and Amodei was the only one at the press conference who insisted “the technology has real risks”.
Sceptics. ABC News noted the accord is thin on detail, saying the signatories “believe” companies “should implement” controls; OpenAI and Anthropic had already committed to third-party evaluators, and it was not clear whether the accord asks them to do anything different. An opinion column in Newsweek pointed out that the “almost constitutional” document contains 308 words of “vague platitudes and loose commitments”, with legal oversight deferred. Asked why stricter guardrails weren’t needed, Trump said the AI chief executives “love” America and are “the smartest people in the world”.
Things to keep in mind
- No enforcement, no penalties. The accord does not say what happens if a company falls short, or who checks.
- No named auditors or timetable. It does not name the outside auditors or say when the steps will be in place.
- “SI” is government language. The order covers federal documents and communications. According to the San Francisco Standard, the leading labs have not relabelled their models as SI, and the heads of Google, OpenAI and Anthropic did not use the term at the press conference.
- “Super Intelligence” in the title is Trump’s term for AI; it does not mean the signatories say they have built systems smarter than people.
What it means for ordinary users
Nothing in ChatGPT, Claude, Gemini or similar products changes because of the accord. It shows the US government’s current approach: no new regulator, but companies setting up their own internal controls and outside audits while the government “watches over” them.
The next thing to watch is how the government follows up. On 4 October Trump announced a “Super Intelligence Force”, led by the Director of National Intelligence, to coordinate federal AI work (see Trump creates a “Super Intelligence Force”).