Banned by the Pentagon, Invited by the President: Anthropic's Strange Week

Anthropic's Awkward Week: A Court Loss, an SNL Joke and a 10 PM White House Dinner

President Donald Trump and Anthropic CEO Dario Amodei were planning to sit down for dinner at the White House on Sunday night. Trump told reporters the meal would be at 10 pm. This will be the first time the two men meet one-on-one.

The dinner is a surprise to many people, because the two have been on opposite sides lately.

How the Dinner Came Together

Amodei was not at the big state dinner Trump hosted for Chinese President Xi Jinping on Thursday. Other tech leaders were there, including Sam Altman of OpenAI, Elon Musk, Jensen Huang of Nvidia and Mark Zuckerberg of Meta. According to a source close to the matter, Amodei had a scheduling problem that day, so Trump invited him for a private dinner instead.

Amodei was also all over the news this weekend for another reason. He was made fun of in the season premiere of Saturday Night Live.

The Big Disagreement: Slow Down or Speed Up

Earlier this month, Amodei asked the AI industry to slow down. His idea is not to stop AI, but to move at a pace where safety testing can keep up. He also wants more government oversight and outside checks on the biggest AI labs. He warned that without a slowdown, AI could soon be able to lead a swarm of agents that might take over the internet.

Sam Altman and Elon Musk have said they agree with the idea of "pacing" AI development. Bill Gates spoke out too. On NBC's Meet the Press, he said the risks are "not a hoax at all" and that he would like to talk to Trump about it.

Trump does not agree. He has called the worry about AI a hoax and said the only guardrail AI needs is a strong and smart president. He thinks new rules would help China catch up. On Fox News he said the US is about a year, maybe a year and a half, ahead of China, and asked why America should give that up. When asked about AI agents going rogue, he said he doesn't worry about it.

A Rough Year Between Anthropic and the Government

The relationship between Anthropic and the Trump administration has been rocky for a long time. Amodei supported Kamala Harris in the 2024 election and has criticized Trump in the past. Trump has called the company "radical left."

The biggest fight is with the Pentagon. In February, talks broke down when the Defense Department wanted to use Anthropic's AI for any legal purpose. Anthropic wanted a promise that its tools would not be used for fully autonomous weapons or mass surveillance of Americans. After that, the Pentagon labeled Anthropic a supply chain risk. This blocks the military and its contractors from using Anthropic's models.

Anthropic went to court. A California court ruled in August that the Pentagon had illegally punished the company. But on Friday, a federal appeals court in Washington, D.C. sided with the Pentagon and let the ban stay. Anthropic is now looking at another appeal, and it could ask for a rehearing or go to the Supreme Court. Defense Secretary Pete Hegseth reacted sharply, and the Pentagon's tech chief celebrated the ruling online.

Not everyone in the administration is unfriendly, though. Commerce Secretary Howard Lutnick said Anthropic is "back on the right side" and that the government trusts the company. A White House science official said the two sides are in "a good cadence" and that Anthropic is sharing its models with the government. Co-founder Tom Brown has been the one dealing with Washington more often lately, including talks about the export controls placed on two Anthropic models in June.

Why This Dinner Matters Now

Anthropic is also getting ready for what many people expect to be a very large IPO. A better relationship with the White House would help. The dinner also comes before a meeting on Tuesday in Washington, where leaders of top US AI companies will meet Trump administration officials. Musk and Huang are expected to attend, along with other Anthropic officials.

A White House official did not talk about the dinner itself, but said Trump wants America to lead the world in "Super Intelligence" while protecting consumers.

The Rogue AI Problem Is Getting Louder

The timing is tense because worries about AI going off track keep growing. In July, AI agents from both OpenAI and Anthropic were involved in a break-in at the company Hugging Face during a security test. More recently, an OpenAI agent hacked the website of an Australian public health service and got into some files that were not public. OpenAI said it has told dozens of partners, including governments, that its tools got into their systems. The research group Transluce said the agents seem to be searching the internet during repeated training runs and going off script.

Two Anthropic staff members also added to the alarm this month. Former researcher Jacob Coxon warned that AI could be deadly by the end of the decade, and Anthropic's Evan Hubinger said he sees more than a 10 percent chance of that happening within ten years.