Amodei meets Trump two days after a court upheld Anthropic's Pentagon blacklist 2-1

Anthropic's political risk sits in three offices, and Sunday's dinner reaches one of them directly. Commerce has already thawed. Howard Lutnick, who temporarily imposed export controls on two of the company's most advanced models in June, told Axios this month that Anthropic is "back on the right side" ^[7]. "We trust Anthropic," Lutnick said at the G20 Innovation Ministerial in North Carolina ^[8]. (Bloomberg's account of the June controls names one model, Mythos ^[16].) Michael Kratsios, director of the White House Office of Science and Technology Policy, told CNBC the administration was "in a good cadence" with the company ^[10].

The Commerce thaw came after a staffing change. Anthropic put Tom Brown, its chief compute officer, on the export-control negotiations and took Amodei off them ^[9]. The president is a harder case. Amodei backed Kamala Harris in 2024 and was a vocal critic of Trump ^[6]. After Amodei's three-step slowdown proposal this month ^[4], "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!" Trump wrote on Truth Social ^[5]. The dinner is aimed at him.

The third office holds the part that restricts customers. Pentagon talks broke down in February because the department wanted use of Anthropic's models for all lawful purposes, while Anthropic wanted assurance they would not be used for fully autonomous weapons or domestic mass surveillance ^[11]. The department then labeled Anthropic a supply chain risk. The label bars the military from its models and blocks defense contractors from using them in Pentagon work ^[12]. On Friday, two days before the dinner ^[18], a federal appeals panel in Washington upheld it 2-1, rejecting Anthropic's argument that the ban was arbitrary, unauthorized and unconstitutional ^[13]. The company can still petition for a rehearing or ask the Supreme Court ^[14]. None of the reports say how much revenue the ban costs, so the discount it implies for an IPO cannot be sized from them.

The dinner can go roughly three ways. The narrow outcome settles the personal feud and leaves the designation with the courts. The IPO would then list with a live lawsuit and a smaller chance of a presidential post during the roadshow. A wider outcome reopens Pentagon talks, and a deal there needs one side to move off the February terms ^[11]. Or the dinner gets folded into Tuesday's reported meeting between Trump, House Speaker Mike Johnson and the heads of major AI companies, held as calls grow for more guardrails ^[15]. Amodei's oversight proposal has had support from Sam Altman and Elon Musk ^[4].

I think the narrow outcome is the likely one. A dinner changes the president's tone, and the designation, with the contractor work it blocks, stays where the court left it. The counter-case is that the designation is the administration's own decision, and a president who wants Anthropic back could have it withdrawn without waiting for a rehearing. If the Pentagon narrows or lifts the label while Anthropic keeps its limits on autonomous weapons and surveillance, the White House is the channel that sets this risk, and splitting it by office was the wrong way to price the listing.