On a late September day in Washington, the leaders of the companies building some of the most consequential technology of this era gathered with the President and agreed, in broad outline, to a set of voluntary safety commitments for artificial intelligence. The pledges, which include working with independent auditors and trying to keep AI tools from reaching technical systems in unintended ways, are modest in their particulars and unclear in their enforcement. Behind the good humor of the room lies an older human question, how a society asks its most powerful makers to police themselves befor
Trump Asks Anthropic's Amodei to 'Be Careful' as Tech Firms Agree to Voluntary AI Safety Standards
Almost like a constitution, in a way — but voluntary
Mark
So what actually came out of this meeting?
Mimi
A voluntary agreement. Several AI firms agreed to work with independent auditors and to try to keep their systems from accessing technical systems in unintended ways.
Mark
And Trump called it almost like a constitution.
Mimi
According to Reuters, yes. That's a big comparison for something voluntary.
Luke
It is. A constitution binds people. Nothing reported here says what binds these companies. No penalties, no named auditors.
Mark
Why did the 'be careful' joke to Amodei get so much attention?
Mimi
Because Amodei had called earlier this month for slowing the pace of advanced AI development, after warnings from former Anthropic and OpenAI researchers.
Luke
Worth noting, though, that we only have the joke and its framing. Nothing in the reporting says Trump was responding to that call, or that anyone's position changed.
Mark
Amodei said they can 'win safely' if they work with the President.
Mimi
That's him describing a cooperative path, not necessarily endorsing the agreement's strength.
Luke
Right. And the 10-person board is only something Trump said he would consider. No members, no timeline. That's an idea, not an institution.
Le Pouls
- A White House gathering of AI and tech executives quickly turned into a stage for public jokes, even as the stakes of the technology remain high.
- Anthropic's Dario Amodei, days after calling for a slowdown in advanced AI development, was teased by the President to 'be careful' in front of cameras.
- OpenAI, Anthropic, Nvidia, SpaceX and others agreed to voluntary standards involving independent audits and limits on unintended system access.
- Trump floated a 10-person AI safety board without naming members, leaving the path from pledge to real oversight undefined.
- Enforceability remains the open question: no clear mechanism yet shows what happens if a signatory falls short.
On a late September day in Washington, the leaders of the companies building some of the most consequential technology of this era gathered with the President and agreed, in broad outline, to a set of voluntary safety commitments for artificial intelligence. The pledges, which include working with independent auditors and trying to keep AI tools from reaching technical systems in unintended ways, are modest in their particulars and unclear in their enforcement. Behind the good humor of the room lies an older human question, how a society asks its most powerful makers to police themselves before it decides whether to ask more of them.
At the White House, President Donald Trump stood before reporters after hosting leaders from OpenAI, Anthropic, Nvidia and SpaceX, with Google's Sundar Pichai and Meta's Mark Zuckerberg also in the room. The occasion was the announcement that these companies had agreed to voluntary safety standards for artificial intelligence.
The mood was light at times. Trump introduced Anthropic's Dario Amodei by saying whatever he says "is okay," then jokingly told him to "be careful." The teasing had context: earlier this month Amodei had called for slowing the development of advanced AI models, amid warnings from former researchers at Anthropic and OpenAI about the technology's destructive potential. Trump also joked about Pichai, praising his life away from daily reporters and urging the room to remember his first name.
Amodei offered a more serious note in a separate video, saying AI brings incredible benefits along with large risks, and that the industry could "win safely" by working with the President. Trump called the agreement "almost like a constitution, in a way."
The terms are sparse. The companies will work with independent auditors to check that their systems behave as intended and will try to keep their tools from accessing technical systems in unintended ways. Zuckerberg said they would build robust internal controls. Trump also said he would consider a 10-person AI safety board, though he did not say who might serve.
Much remains unsettled. The standards are voluntary, the auditors are unnamed, and it is unclear what follows if a company falls short. Those answers will determine whether the agreement becomes real oversight or remains a promise.
After a meeting at the White House with some of the most powerful figures in technology, President Donald Trump stood before reporters and announced that a group of AI companies had agreed to establish voluntary safety standards for artificial intelligence. He had hosted leaders from OpenAI, Anthropic, Nvidia and SpaceX. Google chief Sundar Pichai and Meta CEO Mark Zuckerberg were also among those in attendance.
The moment that drew attention came as Trump introduced Anthropic cofounder and CEO Dario Amodei to the press. Trump said that whatever Amodei says "is okay," then jokingly asked him to "be careful." The quip carried some context. Earlier this month, Amodei had called for a slowdown in the pace at which advanced AI models are developed, a statement that came amid warnings from former Anthropic and OpenAI researchers about the destructive potential of the technology.
Trump's humor extended to Pichai. Praising his success at the head of Google, Trump said the executive has a "great life" because he does not have to face reporters every day. Then he called him "a monster" whom "nobody knows," and repeatedly urged the room to remember the man's first name. "All you have to know is Sundar. Sundar is his name," Trump said. Pichai answered in kind, hoping for "a better tagline than I'm a monster!"
Amodei, in a separate video, struck a more earnest note. He said AI has "incredible benefits" but also carries some big risks. "If we do this right, we work with the President and everyone here, we can win safely," he said.
As for the agreement itself, Trump described it, according to Reuters, as "almost like a constitution, in a way." Its terms, as reported, are modest in specificity. The companies agreed to work with independent auditors to check whether their AI systems are working as intended. They also said they would try to ensure their tools do not "hack or access technical systems in unintended ways." Zuckerberg, who also spoke to reporters, said the signatories would develop "robust internal controls" for their AI systems.
Trump added that he would consider setting up a 10-person board for AI safety. He did not say who might sit on it.
What the reporting leaves open is considerable. The standards are voluntary, and the coverage does not say how compliance would be checked beyond the independent auditors, who they would be, or what happens if a company falls short. The proposed board is only under consideration, with no membership named. Those details, and whether the agreement has any enforcement behind it, are what to watch next.
Citations marquantes
If we do this right, we work with the President and everyone here, we can win safely.— Dario Amodei, Anthropic CEO
The companies would develop robust internal controls for their AI systems.— Mark Zuckerberg, Meta CEO
Envie de l'histoire complète ?
Lire l'original sur CNBC TV18 ↗
Technology