Amodei sits down to dinner with Trump while Anthropic litigates against the Pentagon over military use of its AI

By Zendoric · September 28, 2026.

Dario Amodei, Anthropic's chief executive, will have dinner with Donald Trump at the White House, as first reported by Axios and later confirmed by TechCrunch citing a source close to the executive's plans. It will be their first face-to-face meeting, and it comes at a particularly charged moment: Amodei has publicly argued that AI development should slow down or, at the very least, proceed with far more caution. Trump, without providing evidence, has called those criticisms a "Democrat hoax" and has even proposed renaming the technology "superintelligence".

The dinner does not come out of nowhere. According to the source, the Pentagon this year designated Anthropic a "supply chain risk" after the company tried to impose restrictions on the use of its own technology; Anthropic is challenging that decision in court. At the same time, other Administration officials have been more receptive to the company, which sketches a divided internal front over how to treat the lab that has insisted most on the need for safety guardrails.

The context has thickened in recent days. The source also notes that the White House has asked OpenAI and Anthropic to withhold their newest models from British evaluators until US authorities review them first, citing cybersecurity risks, a move that, should its scope be confirmed, strains coordination with allies in the middle of the AI race. And Anthropic is preparing, ahead of a possible IPO, a supervoting structure that would give Amodei and the other six co-founders 50.1% of the votes on most corporate decisions.

Beyond the regulatory arena, the debate has also seeped into popular culture: Saturday Night Live devoted a sketch to parodying Amodei's apocalyptic warnings about AI, a sign that the dispute over the risks has stopped being a technical niche matter and become late-night material. That same day, Bill Gates called for mandatory government oversight of AI safety, an institutional route that contrasts with the dinner-table diplomacy Amodei is pursuing.

Our reading: this dinner is not an armistice, it is a photo op. Trump has not moved his position on AI safety —he still treats it as a rhetorical excuse from the opposition, not as a real risk— and Anthropic remains in court fighting the Pentagon itself over the very issue behind Amodei's caution. What changes is the channel: when the regulatory route is blocked or hostile, frontier labs fall back on a personal relationship with the executive branch. Broadly speaking, the brakes are being called for by those building the technology; the pedal is still controlled by politics.

That shift in where decisions are made —toward personal proximity to power, rather than toward evidence-based regulation— is precisely the kind of short-term risk that worries us: not runaway superintelligence, but the concentration of the power to decide about it in very few hands, whether those of a handful of founders with supervoting rights ahead of an IPO or those of whoever manages to secure a seat at a White House dinner. In the long run we remain convinced that AI can lead us to a society of unprecedented abundance and health; but that future is settled, in large part, by governance decisions taken today in offices and dining rooms, not only in labs.