An Anthropic Essay About Slowing Down AI Just Got Attacked by Both the White House and Beijing

On Saturday, Anthropic chief executive Dario Amodei published a 3,800 word essay asking the AI industry to slow itself down. By Monday, the American president had called it a sick conspiracy and the Chinese government had called it a Cold War tactic. Two governments that agree on almost nothing managed to agree that Amodei was wrong, for reasons that had nothing to do with each other. In between, Anthropic quietly told its own investors that business had never been better. Away from all of it, India’s cricketers put together the kind of week that needs no argument at all.

A weekend essay warns AI swarms could take over the internet

Amodei’s essay, titled “We Must Pace the Frontier,” argues that AI companies should deliberately slow the rate at which they push model capability forward, by a year or two, so that safety research has time to catch up. His central warning is specific: he thinks a swarm of AI agents could take over large parts of the internet with a persistent botnet within six to twelve months, causing hundreds of billions of dollars in damage, unless labs and governments act first.

His proposal comes in three steps. The first, which Anthropic says it will do immediately and alone, is giving outside evaluators such as METR the kind of ongoing, employee level access it would take to actually check a company’s safety claims: office access, laptops, permission to publish findings without the company editing them first. The second step asks competing labs to adopt shared safety standards together. The third asks for coordination that reaches authoritarian governments too, which is the part nobody expects to happen soon.

Within hours, OpenAI’s Sam Altman posted that he agreed Amodei was right and that OpenAI would give outside evaluators similar access. Elon Musk endorsed the call as well. For a few days, three men who compete directly for the same customers agreed in public on something. That did not last.

The president calls it a “sick conspiracy”

On Monday, President Trump used his own account to reject the entire idea. “There is a SICK conspiracy going on against AI and Data Centers, and the only one that is happy about it is China,” he wrote. “WHOEVER WINS AI, WINS!” He argued that the only guardrail AI needs is a strong, smart president, and singled out Amodei by name, accusing him of “pretending to be a perfect little angel” after years of running an AI company. Notably, Trump did not attack Altman or Musk by name, even though both had endorsed the same slowdown call days earlier.

The administration has spent 2026 pushing hard in the opposite direction from Amodei’s proposal, backing rapid data center construction across the country as part of a strategy to keep the United States ahead of China on AI capability. From that view, any voluntary slowdown looks like giving up ground for no reason.

Beijing calls it a Cold War tactic instead

China’s objection is the mirror image of Trump’s, and just as blunt. Asked about the slowdown push at a Monday briefing, foreign ministry spokesperson Guo Jiakun said “fearmongering, confrontation and vicious competition will only disrupt the process of global AI governance and serve the interests of no one.” State media went further. The Global Times called the proposal, which also touches export controls on AI chips to China, “a Cold War playbook for the AI sector,” and an editorial in China Daily argued that after chip and equipment restrictions failed to slow Chinese AI progress, the same goal was now being pursued through calls to restrict software and capability instead.

So the White House says the slowdown call helps China, and the Chinese government says the same call is designed to hurt China. Both cannot be right about the intent behind one essay. What is true is that neither government wants anyone pacing anything, which tells you more about the state of the AI race than either statement does on its own.

The numbers behind why everyone is fighting over this

While the argument played out in public, Anthropic gave a smaller group of investors a look at why so much rides on the outcome. The company said it expects a second consecutive quarter of positive adjusted operating income, on second quarter revenue of $11.5 billion, up from $4.73 billion in the first quarter of 2026, with gross margins above 80 percent before accounting for the cost of training its own models. Annualized revenue reportedly reached $65 billion by the end of July.

Anthropic has also picked Nasdaq for a planned initial public offering, with an October listing window under discussion. Some reports put a potential valuation as high as $2 trillion, more than double the roughly $965 billion the company was valued at in a private funding round this past May. None of that is settled yet, but it explains why a debate about pacing frontier AI is not really a debate about physics or safety research alone. It is also a debate about who gets to own the most valuable company in the world, and how fast.

India’s cricketers made a record with no asterisk on it

Away from the argument, India’s women’s cricket team gave the country something to celebrate without any caveats attached. In Dubai on Saturday, they beat defending champions Sri Lanka by 72 runs to win a record extending eighth Women’s Asia Cup title.

Openers Shafali Verma and Shree Charani built the foundation: Verma reached fifty off just 24 balls, the fastest half century in Women’s T20 Asia Cup history, and finished with 68 off 39 balls. Smriti Mandhana added 59 off 39 balls at the other end. Their opening stand of 129 is the highest partnership for any wicket in the tournament’s history, and it carried India to 183 for 5. Sri Lanka never got close, bowled out for 111 as spinner Shree Charani took four wickets for 20 runs.

Eight titles out of the tournaments India has entered is not a fluke, it is a program that keeps producing players who can do this on demand. Nobody needs to explain away a single number in that scoreline.