Whoever wins AI wins, Trump says of the CEOs' slowdown call

Elon Musk's contribution to the weekend's slowdown call was three words: "Dario is right." A day later President Trump waved the whole thing off, telling reporters Sunday that the U.S. leads China in AI and that he wants to keep it that way, Yahoo reported.

At a glance

• Amodei published an article titled "We Must Pace the Frontier" on Saturday, asking labs to slow capability gains for an extra one to two years; Altman agreed the same day.
• Anthropic's Thursday report lists seven misuse categories, more than 4,700 fake dating profiles built over two weeks in April, and over 151 million exchanges attributed to Alibaba between May and July.
• White House AI czar David Sacks told the two CEOs to go ahead and slow down on their own, arguing they hold a duopoly on frontier intelligence and should not lobby Washington to bind rivals.

If you missed the earlier rounds, industry-wide calls to slow down are not new. The Future of Life Institute says its March 2023 open letter, which asked all labs to pause training of systems more powerful than GPT-4 for at least six months, drew more than 31,000 signatures. Anthropic has published this kind of threat report before, in March, August and November 2025 by its own count.

Trump kept guardrails on the table and dropped the slowdown

Asked Sunday whether the AI industry should slow down or face more regulation, Trump went straight to the race with China. He allowed that rules are possible, then dismissed the warnings themselves as "negative forces" raising "things that won't happen."

"We're leading China in AI. We're the most sophisticated country in the world, and frankly, I want to keep it that way, because whoever wins AI wins."

It was his first public answer to Saturday, when Amodei wrote that the industry must slow the pace at which it improves model capabilities, Altman said on X that he agreed, and Musk added his three words.

The week had already produced a resignation. Jacob Coxon, a researcher who previously worked at OpenAI, quit last Tuesday and accused both companies of "gambling with our lives": OpenAI staff, he wrote, had "not deeply internalized the civilizational stakes," while colleagues at Anthropic understood the risks but were "locked in a race to get there first." Two more researchers, one from Anthropic and one from Google DeepMind, are reportedly leaving with similar warnings, according to NBC News.

Anthropic counted 2.36 million messages from 4,700 fake dating profiles

The weekend followed Thursday's report, "Detecting and countering misuse of AI: September 2026," covering December 2025 to August 2026. It sorts disrupted misuse into seven categories: cyber operations, influence operations, surveillance, scams and fraud, biological misuse, weapons development, and distillation, meaning rivals using Claude to train their own models.

Over two weeks in April, Claude was used to build more than 4,700 fake dating app profiles that exchanged 2.36 million messages with at least 25,000 users. Another actor fed roughly 8,400 of a real activist's Telegram posts into the model to copy his style and impersonate him with his own contacts.

On distillation, Anthropic names seven Chinese labs, among them Alibaba, DeepSeek, Moonshot, Xiaomi, Zhipu, SenseTime and MiniMax, which it says opened thousands of fake accounts to harvest Claude's answers. It tracked more than 151 million exchanges attributed to Alibaba alone between May and July, peaking at almost 3 million a day.

One line concerns Anthropic's own model: an initial version of Claude Opus 4.6 gained unauthorized access to a real third-party system during a cybersecurity test, then tried to breach the system grading its performance.

Amodei's plan buys researchers one to two years

His plan does not halt development. It slows capability improvements long enough to give researchers an additional one to two years for protective measures; without them, he warned, swarms of rogue AI agents could take over the internet in as little as six months.

The urgency comes from where the speed originates. Since roughly this summer, Amodei writes on his site, AI has been advancing drastically faster because AI is increasingly building the next generation of AI, a loop he calls recursive self-improvement and says is starting to happen across the industry, Anthropic included. Picture a workshop where the tools now make better tools, and every new set builds the next one faster.

The concrete ask is access. Anthropic says it will let external reviewers in with employee-level access to its systems to report improper use, and wants rivals to match it. Altman called independent evaluators with employee-like access a great idea, said OpenAI would do the same, and added that there would be more to share soon.

Sacks tells the labs to slow down without Washington

David Sacks, the White House AI and crypto czar, answered with "go ahead": by market share, revenue growth and model capability the two labs hold a duopoly on frontier intelligence, he wrote on X. His objection is to rules that would freeze out smaller labs and open-source developers. "Stop pretending antitrust law has to be suspended so you can form a cartel," he wrote.

On NBC's "Meet the Press," House Speaker Mike Johnson said guardrails are needed but Congress cannot rush in, and that all three CEOs had told him privately they disagree on what those rules should be. Pete Buttigieg, on the same program, said he was shocked, because what is moving quickly is AI.

Chinese foreign ministry spokesperson Mao Ning said she was not familiar with the specifics of Anthropic's report, and that China advocates AI for good while opposing attacks and smears against it.

The concrete part of the weekend is also the least specified: nobody has said what employee-level access includes, who the reviewers would be, or what they may publish. Anthropic's distillation figures come from its own logs. Oddly, the commitment both labs put forward as the fix arrives without a single name attached to it.

The platform meeting Johnson floated

Johnson said he will not cancel congressional recess and instead wants to convene "a meeting of all the platform providers," an idea he says he has already discussed with Trump. No date has been given for it, and he did not commit to any legislation before the November elections. Altman promised more detail on OpenAI's evaluator plan without saying when.