Trump Declares AI Giants Are "Self-Policing" After White House Safety Accord

US President Donald Trump announced Tuesday evening that a group of leading artificial intelligence executives had signed a non-legally binding agreement to police their own industry, saying he was witnessing "tremendous self-policing" by companies that "understand that they have to self-police." The accord was unveiled at an impromptu press conference outside the West Wing and posted late Tuesday on Trump's Truth Social account.

The signing capped a day of meetings at the White House between Trump, House Speaker Mike Johnson and a roster of chief executives from technology companies collectively spending hundreds of billions of dollars on AI infrastructure — including the data centres now sprouting in communities across the United States, which have drawn grassroots opposition that spans party lines. Trump said the executives had agreed to help ease that backlash by channeling more money toward local schools and working to bring down energy costs for residents living near the sites.

Among those gathered were CEOs who have publicly urged the industry to slow its pace in the wake of a string of high-profile episodes in which AI systems have "gone rogue." That tension inside the room was on display throughout the day, even as the industry locked in a common framework. Related France 24 reporting documented one firm, OpenAI, cancelling the release of a new model over safety concerns just hours before the accord was announced.

The document itself sets out four voluntary steps. Companies commit to building "robust internal controls," teaming up with an "independent external auditor" to determine whether those controls are genuinely effective, and creating a dedicated committee within each board of directors to review the findings produced by both internal and external auditors. The arrangement stops short of creating a regulator, but it explicitly preserves the option of one: the accord states that "over time, it may make sense to codify these steps into laws and regulations."

Trump characterised the deal as "morally binding" rather than enforceable in law, and said roughly a dozen people would eventually sit on a committee charged with "watch over the whole enterprise." He added that after consulting with industry leaders, he would name a single individual to oversee implementation of the agreement in the days ahead — a detail that suggests the White House intends to keep a running tab on compliance even without statutory teeth.

Dario Amodei, whose comments were cited during the discussions, offered a more guarded assessment, noting that "the mechanism, how we address those risks is still under discussion." His remarks captured the accord's central ambiguity: the signatories agree that safety reviews matter, but not yet on who sets the standard, who audits the auditors, or what happens when a company decides the commercial cost of compliance is too high.

Trump, for his part, has consistently rejected any role for the government in slowing the technology. Earlier Tuesday, he promised that the US would not limit AI development and would not back calls for "guardrails" or other restrictions. Speaking at the unveiling of America.gov, a chatbot designed to help Americans navigate federal services, he vowed: "I will never stifle the growth of a technology that will be bigger than the industrial revolution." Trump added: "We can't stifle it, and we're leading by a lot."

The America.gov launch was the centrepiece of a daylong Washington event at which administration officials toasted what they described as a "Golden Age" powered by what Trump insists on calling "super intelligence" rather than artificial intelligence. Secretary of State Marco Rubio used the occasion to preview a practical upgrade: applying for a first-time passport entirely online, without the in-person visit to a processing centre that applicants are currently required to make.

Economic arguments ran through much of Trump's praise for the industry. He credited AI developers with a technology he said will deliver "unbelievable growth," pointing to the data centre buildout and the computer equipment that accompanies it as accelerants of the wider economy. That spending, he said, has buoyed the stock market, swelled the investment accounts of higher-income Americans and lifted their consumer spending.

On the data centre question specifically, Trump forecast a shift in local politics: "You're going to see, data centres are going to be very popular." Should communities continue to resist the facilities, he warned, the companies "will be forced to go overseas or other locations" — a warning that carries obvious implications for US investment, electricity demand and regional employment.

The accord's practical test will come in the months ahead. Signing a pledge changes nothing on its own; the scrutiny will fall on whether the promised internal controls survive commercial pressure, whether outside auditors enjoy enough independence to issue unflattering verdicts, and whether board-level committees take the reports seriously rather than rubber-stamping them. Each of those questions is precisely the kind of dispute that Amodei's remark about the mechanism remaining "under discussion" was meant to keep open.

For now, the industry has bought itself time and a measure of political cover. The phrase "morally binding" is doing a great deal of work in a document that creates no penalties, no reporting obligations and no enforcement body — only a promise, a forthcoming White House point person, and a set of four steps that could eventually be written into law if the voluntary model holds.

Trump, meanwhile, is betting that speed and safety are not rivals. The CEOs who joined him at the table appear less convinced. But after a day of mutual accommodation, both camps left Washington with the same message: the argument over how much AI should regulate itself has been postponed, not settled.