Trump calls AI warnings a 'sick conspiracy' — and his own adviser says the labs could just stop

The White House escalated this week from indifference about AI risk to open attack on the people raising it. On Sunday, President Trump told reporters in Doonbeg, Ireland, that the calls for a slowdown came from "negative forces" bringing up "things that won't happen." By Monday he was posting on Truth Social that there is a "SICK conspiracy going on against AI and data centers, and the only one that is happy about it is China" — and naming Anthropic chief executive Dario Amodei. The counter-argument he did not make arrived within hours from David Sacks, co-chair of his own Council of Advisors on Science and Technology: worried labs, Sacks wrote, can slow down without anyone's permission.

The attack lands on the messenger

In Doonbeg, after the Irish Open, Trump conceded the principle and emptied it of content. "We can put guardrails, we can do this and that, but I think you have a lot of negative forces that are bringing it up that shouldn't be bringing it up, and they're bringing up things that won't happen," he said. He acknowledged that some regulation was needed and gave no detail. What he wanted, he said, was to keep the pace: "Frankly, I want to keep it that way because whoever wins AI wins."

Monday's Truth Social post went further. His administration, Trump wrote, had already stopped AI companies from doing "bad, or potentially bad" things, referring to Amodei "who is now pretending to be a 'perfect little angel' — and we will continue to do so." He warned "conspiracy theorists, treasonists, traitors, and leakers," and closed on the constitutional theory: "The only control or 'guardrails' that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!"

That is an attack on the credibility of a sitting lab chief executive, not a rebuttal of his argument. Amodei's essay, published on Saturday, warns that AI agents could "take over the entire internet" within six to 12 months. The president's reply to a man who has argued for years about catastrophic risk was to call him an angel impersonator. The earlier pieces of the sequence are on the record: the resignation of Anthropic researcher Jacob Coxon, who said staff were "genuinely frightened", and Trump's own September 11 dismissal, that AI would be fine and the country would have "a little gear".

A counterpunch from inside the president's own council

Sacks was blunter about the labs than the president was. "Stop pretending you need anyone else's permission," he wrote on X. Then: "So go ahead and pace the frontier. You are the ones setting it... Demanding your preferred regulatory framework as the price of that will look like blackmail of the public and the political system. So just do it."

The argument is checkable, and that is what makes it awkward. Amodei's plan has three steps, and only the first — embedded third-party evaluators with employee-like access inside each lab — is presented as something a company can do alone; Anthropic says it is committing to it now. The other two require democratic governments to coordinate, which means asking the public for something. Sacks's point is that nothing technical prevents a lab from starting today: no statute, no waiver, no summit. The obvious reply is that a pause by one company hands the frontier to whoever declines to pause — but that is an argument about competitive consequences, not about permission.

Congress is split along the same line

House Speaker Mike Johnson told CNN's State of the Union that an emergency session on AI would lose the race to China: "If Congress just races in and does some sort of emergency session to try to regulate AI, we will lose the race to China, and that is a threat to every single American." House Minority Leader Hakeem Jeffries, on ABC's This Week, argued the opposite, that the United States should slow "the pace of development in order to protect the American people and ensure that AI is proceeding safely."

Who	What they said	When	Where
Trump	oversight demands are a "SICK conspiracy"	Sep 14	Truth Social
Sacks, PCAST co-chair	labs should "stop pretending you need anyone else's permission"	Sep 13	X
Johnson, House Speaker	an emergency session "will lose the race to China"	Sep 13	CNN
Jeffries, Minority Leader	slow "the pace of development"	Sep 13	ABC

A framework already exists on paper. The bipartisan Frontier Act, introduced in July by House Democrats and Republicans, would create a national safety and oversight regime. Its co-author, Representative Lori Trahan, wrote on X this week that "safety researchers are resigning, powerful AI models are breaking out of their labs, and companies are racing ahead anyway," adding: "It's past time for Congress to get off the sidelines and do its job."

The market moved before the diplomacy did

Markets did not wait. AI-linked stocks plunged worldwide on Monday, Reuters reported, and chips led the decline: the Nasdaq 100 slid 1.2% to a six-week low in early trading before paring losses to 0.4%, and the Philadelphia semiconductor index dropped 5.2%, with Nvidia down 3%, AMD 4.5% and Micron 5.4%. Lam Research and Applied Materials each lost more than 6%. Europe's tech sector fell 2.2%, dragged by ASML's 6% decline; in Asia, SoftBank plunged more than 10%. Sam Altman also said OpenAI will not proceed with an IPO this year, citing safety concerns.

The repricing is the week's clearest measurable event. An industry that says it is frightened of its own products is still shipping them: OpenAI's GPT-6 Astra rolled out at the start of the month, days before Amodei's essay, and the essay's own first step carries no date, no metric and no penalty for failure. Investors marked the companies down for the possibility of a slowdown while the labs kept operating on the assumption of none. Steve Sosnick of Interactive Brokers put the dependency plainly: "we've been running hot based on AI spending."

What would settle it

The president's closing line is a governing philosophy rather than a control mechanism. "The only control or 'guardrails' that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT" contains no threshold, no audit, no incident-reporting duty and no test that could fail; nothing in it can be checked against a model's behavior, which means it cannot be argued with on the evidence. Sacks's position has the opposite property — it is falsifiable within weeks by a lab that simply begins.

What would settle the argument is not another round of posts. It is published post-training and evaluation results, third-party audits with real access to training runs and weights, and a lab that actually pauses a frontier run, with a date attached so the pause can be verified rather than announced. Until at least one of those exists, both camps are asking to be believed: Amodei on the claim that his own industry cannot stop, and the president on the claim that he is all the stopping it needs.