Amodei calls for slowing down on artificial intelligence, Trump rules out any brake

In the morning edition the Tech section was taken up by European space. The afternoon brings a debate that opened over the weekend and received a political response on Monday: to what extent the artificial intelligence industry should regulate itself, and at what speed.

Dario Amodei, chief executive of Anthropic, has asked companies in the sector to slow down to allow safety to catch up. His forecast is precise in timing and subject: within six to twelve months an artificial intelligence system could be capable of directing a swarm of programs able to take control of much of the internet. It should be stated clearly that this is the forward-looking assessment of a company executive, not a published experimental result: we report it as the position of the person who expressed it.

The concrete proposal

The operative element of the request is the presence of external reviewers inside major artificial intelligence companies, with the right to publish their own findings. This is the point that distinguishes a proposal for oversight from a statement of principle: it is not enough for someone to verify, they must be able to make public what they find.

Sam Altman, of OpenAI, has stated that his company agrees with the proposal. The consensus between the two leading competitors in the sector shifts the question elsewhere: who should set the rules.

The political response

The director of the White House’s National Economic Council, Kevin Hassett, said he believes the sector’s leaders are doing their part. President Donald Trump, asked about the weekend’s warnings, ruled out any slowdown: in his account the United States is ahead of China in artificial intelligence and intends to keep that position, because in his view whoever prevails in this field prevails generally. We report the position in indirect form: the original wording exceeds the length limit this newspaper applies to quotations, in the source language as in translation.

The resulting picture is unusual: the companies developing the systems are asking for a brake, while the administration that should regulate them rejects it in the name of competition with Beijing.

The financial calendar says something else

In the background is a timeline that concerns money. Sam Altman told Fortune that OpenAI will wait until 2027 to begin selling shares to Wall Street investors, linking the delay to a focus on safety. Axios reports on September 14 that the company was leaning toward that date and appears to be sticking with it; Anthropic, by contrast, would likely still be aiming for a listing in 2026.

The paradox is visible: the company whose chief is publicly asking the industry to slow down is the one aiming for the stock market first. The two things do not necessarily contradict each other — going public does not mean accelerating development — but they should be noted together.

Associated Press also reports that a researcher left Anthropic over doubts about responsible conduct in artificial intelligence development. We do not name them: this newspaper’s rules exclude coverage of individuals outside of a public role.

What we don’t know

We don’t know in what setting Amodei made the proposal, nor whether a document exists detailing it. We don’t know whether other companies in the sector, besides OpenAI, have accepted or rejected it. We don’t know on what technical analysis the six-to-twelve-month estimate is based: the available sources report the statement, not the methodology behind it. None of the risk claims reported in this piece has been independently verified by this newspaper; they are the positions of identified parties, and should be read only as such. The only dated deadline is the financial one: 2026 for Anthropic, 2027 for OpenAI.

Sources: WORLD Radio; Everything Briefing; WFDD / NPR; Associated Press; Axios.