Dario Amodei did not arrive at Sunday's dinner with President Donald Trump carrying an easy relationship to repair. The Anthropic chief executive had spent months on the opposite side of the administration in one of Washington's ugliest AI fights. The Pentagon characterized Anthropic's issue as a supply chain problem, and the Commerce Department restricted some exports associated with one of the company's top models.

Yet after the two men met privately for the first time, the president's public tone shifted almost immediately. By the time he spoke to reporters, Trump was talking about the Anthropic CEO very differently.

Trump's new tune

Asked whether he trusted OpenAI CEO Sam Altman, Trump laughed at the premise, then told reporters: "I like him. Do I trust him? Who do I trust?" He went on to group Altman and Amodei together as unusually smart people working at the center of the AI boom.

The president said the dinner changed his view of the Anthropic CEO outright. "I liked him and his wife a lot," Trump said, calling Amodei a "very smart guy" who was not the person he had built up in his head. He shared the remarks publicly, including in a post on Truth Social.

Anthropic and Washington had been at loggerheads throughout the year over how the US government should be allowed to use the company's technology. Amodei has stood out as an executive unusually forthcoming about what he does not want his company's models used for.

Trump has taken a far simpler view of the competition: America is ahead, and he does not want rules that make staying ahead harder. "If you're going to seek regulation, they can put you out of business," Trump said. In his view, the United States is outperforming its rivals in AI and should not voluntarily make things easier for everyone else pursuing the technology.

In September, Amodei said the tech industry needs to take its foot off the pedal. Anthropic also made its model permanently available to outside evaluators to verify whether the company had fulfilled its public safety promises. To Trump, moves like that amount to regulatory bait.

But Amodei's position is not "stop AI." It is stranger — and, frankly, more hawkish. He wants democracies to dominate advanced AI and has pushed an "entente" strategy under which allied democratic countries hold the strongest systems, use them for national security and military purposes where needed, and share the economic upside among friendly states.

Trump, meanwhile, has pointed to the government arrangement he said he reached with Intel Corp. (NASDAQ: INTC) and mused, "Maybe I could do that with OpenAI and Anthropic."

From OpenAI defection to a $965 billion valuation

Amodei has spent more than a decade inside the AI industry. He worked at Baidu (NASDAQ: BIDU) between 2014 and 2015, then moved to Google, now part of Alphabet Inc. (NASDAQ: GOOGL, GOOG). In 2016, he joined OpenAI and eventually became its vice president of research.

Then came the split. Amodei and his Daniela Amodei left OpenAI in 2021 to found Anthropic together with several of their colleagues from the organization, driven by differences of opinion over the company's future direction. Anthropic went on to create Claude, which became the strongest competitor to OpenAI's ChatGPT.

During the chaos surrounding Sam Altman's temporary removal as OpenAI CEO, the company's board approached Amodei about taking the job. There was also discussion of combining Anthropic with OpenAI. He said no to both.

By May 2026, Anthropic had reached a valuation of $965 billion, and Forbes valued Amodei's fortune at $15.5 billion. Time had already named him one of the 100 most influential people in the world in 2025; he made the list again in 2026 alongside Daniela, and both were featured in the magazine's list of "Architects of AI."

In his 2024 essay Machines of Loving Grace, Amodei argued that powerful AI could radically improve biology, neuroscience, economic growth, medicine, global stability, work and daily life. He also wrote that people were underestimating both sides of the technology at once: how good it could become, and how ugly things could get if it goes wrong.

Medicine is personal for him. Amodei has said he lost his father just several years before sofosbuvir, a drug used to treat hepatitis C, came to exist — an experience that shaped his conviction that AI could shorten the time required for medical discoveries and bring treatments to people.

The race Trump wants to win could turn ugly

In January 2026, Amodei published "The Adolescence of Technology." This time, the focus was not miracle drugs and productivity. He broke the dangers into five broad buckets.

First, AI systems could begin doing things their makers never intended them to do. Anthropic's own tests have already yielded models that engage in deception, blackmail and scheming in artificial settings.

Second, Amodei has been particularly concerned about bioweapons, and his fear is unapologetically blunt: a person may no longer need to spend years building expertise when an AI program can walk them through technical procedures that once required it. That narrows the gap between a "person with a terrible idea" and a "person with the ability to do something terrible."

Third, he has cautioned that governments could use advanced AI for surveillance, autonomous weapons and propaganda on a level beyond the capabilities of present technology. He has singled out the Chinese Communist Party as the biggest threat in this regard.

Fourth, Amodei has forecast that AI could eliminate about half of all entry-level white-collar jobs within one to five years. He has also warned that the money generated by AI could concentrate in even fewer hands than during the Gilded Age, perhaps creating fortunes worth trillions of dollars each.

The final risk is that AI could advance biological science far enough to extend the human lifespan and eventually influence human intelligence. Interacting with AI that surpasses us could reshape social and working habits. Should machines become more proficient than mankind in nearly all practical skills, society will ultimately face a difficult dilemma: what should the majority of people do with themselves all day?

Trump believes America can regulate itself right out of first place. Amodei believes America can run itself right into trouble it cannot fix. They did not settle that debate over dinner. What remains to be seen is whether the warmer tone touches anything concrete: the Commerce Department restrictions on one of Anthropic's top models, or the government arrangement Trump said he reached with Intel and has mused about applying to OpenAI and Anthropic.

Source: Cryptopolitan