Why Anthropic’s Dario Amodei angered OpenAI and other AI peers

ByTech Word News

Dario Amodei, CEO of Anthropic, was becoming a nuisance to his peers in the tech industry.

Anthropic and OpenAI, leading AI companies, lobbied the White House to prevent the development and widespread sharing of some AI models because they argued they were too dangerous. But companies like Nvidia and Microsoft saw these AI models as a big business opportunity.

At the end of July, Satya Nadella, the chief executive of Microsoft, and Jensen Huang, then the head of Nvidia, intervened. They emailed drafts of a letter saying those AI models should remain open, according to four people familiar with the discussions. When other technology executives were asked if they would like to sign on, two dozen companies quickly agreed before the letter went public. Over the next week, another 200 companies signed up – including the penalized OpenAI letter.

Dr. Amodei refused.

Dr. For years, Amodei has infuriated his colleagues by warning about the extreme risks of artificial intelligence, even though his company has built some of the most powerful artificial intelligence systems in the world. Frustration has turned to hostility as Anthropic jumps to the forefront of artificial intelligence development, posts stunning revenue growth and moves toward what is likely to be one of the biggest initial public offerings ever later this month.

Along the way, Anthropic, which has cultivated an image as a company that strives to do the right thing, has also earned a reputation for playing hardball with customers and competitors. It cut some off its services and lobbied against free alternatives to its AI from other companies

Over the past year, other tech companies and investors have organized to push back, according to more than 20 people in the tech industry, many of whom spoke on condition of anonymity to preserve their relationships with Anthropic.

They tried to show a united front to counter the warnings of Dr. Amodeia about AI. They relied on their ties to President Trump to keep regulators at bay. They have publicly complained about Anthropic’s business practices. They even spoke to another Anthropic executive about Dr. Amodei scares people with his rhetoric.

But a report on artificial intelligence security by Dr. Amodei, who seems unfazed by criticism from other executives, is breaking through. AI agents attacking websites are becoming more common. And in midterm elections, the threat of job losses due to artificial intelligence and data center construction were significant issues for voters in both parties.

(The New York Times has sued OpenAI and Microsoft, alleging copyright infringement of news content related to AI systems. Both companies have denied the claims.)

Anthropic began with a split. In 2020, a group of OpenAI employees led by Dr. Amodei left the company due to disagreements about how to safely develop AI. Those employees founded Anthropic in San Francisco the next year.

Its founders’ ties to effective altruism — a philosophy where donors seek to measure and maximize the benefits of their philanthropy — formed the foundation of the company, particularly its approach to AI security.

However, Anthropic played second fiddle to OpenAI, making everyday use of AI tools mainstream with the release of ChatGPT in 2022.

That changed in February 2025, when Anthropic surprised competitors with a preview of Claude Code, an AI coding tool. Software engineers started using this tool to generate computer code instead of writing it themselves. Antropic’s sales exploded.

After that, critics say, Anthropic ran into trouble with customers.

A start-up called Windsurf had talks about selling OpenAI and these talks became public in April 2025. Anthropic later cut off Windsurf’s direct access to its models, which the startup used to power its own AI coding tool.

“I think it would be weird if we were to sell Claude to OpenAI,” said Jared Kaplan, one of Anthropic’s co-founders and chief scientist. in the interview in June.

The move crippled Windsurf and potential customers went elsewhere, according to three people familiar with the discussions. Although Windsurf’s negotiations with OpenAI failed, Anthropic had Windsurf cut off.

Windsurf signed a license agreement with Google in July 2025. But the message was delivered to the rest of the startup world: Their business was built on Anthropic models, but to maintain that approach, they needed to stay in the good Anthropic environment.

“In the United States, we’ve somehow convinced ourselves that innovation should be in the hands of a few organizations,” Manos Koukoumidis, chief executive of artificial intelligence startup Oumi, said in an interview.

Last February, about a year after Anthropic introduced it, Claude Code was well on its way Annual revenue of $2.5 billion. In its financial statement, Anthropic disclosed total revenue of $17 billion for its most recent fiscal quarter.

At the beginning of 2025, dr. Amodei has ruffled feathers at Nvidia by supporting US trade restrictions on chip sales to China. That June, Mr. Huang said he disagreed with Dr. Amodei in almost everything.

“They believe that AI is so scary that only they should do it,” Mr. Huang said of Dr. Amodei and Anthropic at a conference in Paris. (Mr. Huang resigned from Nvidia on Thursday under pressure from its board.)

Anthropic replied: “Dario never claimed that ‘only Anthropic’ could create a safe and powerful AI.”

But industry leaders, worried that public sentiment would turn sharply against AI, asked Anthropic to tone down doomsday talk.

Marc Andreessen and Ben Horowitz of venture capital firm Andreessen Horowitz met Tom Brown, co-founder of Anthropic, over the summer. Antropic scared the public with its predictions of extreme job losses, Mr. Andreessen and Mr. Horowitz said, and should cool it, according to three people familiar with the discussion.

Mr. Andreessen then pulled out his phone, where he pulled up one of Dr. Amodeia. The text below Dr. Amodei’s face summarized his prediction of a white-collar bloodbath and mass unemployment. Mr. Brown appeared caught off guard by the confrontation.

Then last June, Anthropic released the Fable, a version of its most advanced model. The company said it would downgrade the model if it thought Fable was being used for AI research or development, but Anthropic wouldn’t tell the user it was downgrading, which angered some researchers.

“It’s a very real sign of concentration of power that companies can engage in such obviously harmful behavior for users and still lead the market,” said Nathan Lambert, co-founder of Trillium Labs, an open-source artificial intelligence nonprofit. in a post on a social network.

Within a day, Anthropic resigned from politics. Now, if it thinks Fable is being used for AI research or development, it will switch to a worse model and tell the user.

“We made the wrong compromise,” Anthropic said at the time.

Anthropic too he said it would store the data of those using its most advanced models for 30 days “to support our security work”. A sign of how paranoid Silicon Valley has become about Anthropic, the change raised concerns that the company could train its models on customer data and then compete with those customers, according to four people familiar with the changes.

In September, Anthropic began storing data for customers on its own cloud services. Danielle Cohen, a spokeswoman for the company, said it treats customer data as confidential and prohibits training about it by default.

In late July, Silicon Valley’s anger toward Anthropic came to a head when Anthropic and OpenAI lobbied the White House to crack down on open-source AI models. Their campaign came after OpenAI said its artificial intelligence had gone rogue and hacked into other companies, including start-up Hugging Face.

So Mr. Huang and Mr. Nadella shared a letter supporting AI with an open scale that makes available the rules that govern the model. That opened the floodgates against Anthropic, according to five people familiar with the discussions. So many companies asked to sign the letter that Microsoft set up an email address to more easily accept signatures for the letter. These included Amazon and Google, both investors in Anthropic.

OpenAI eventually signed the letter, but Dr. Amodei said he didn’t agree with parts of it and Anthropic didn’t sign it. He said in a blog post that all AI models, regardless of how they are developed and shared, “should undergo mandatory security testing.” He said he believes bad actors will exploit open-weight AI models.

Silicon Valley leaders continued to discuss how to present a united front on security issues, mostly keeping Anthropic involved in those discussions. Executives from Nvidia, Microsoft, Google, Meta, Anthropic and OpenAI bounced ideas around in calls and text messages, according to five people familiar with the discussions. Some held informal discussions about a joint statement or commitment to slow the development of artificial intelligence. Others discussed regulatory proposals.

But in September, three days after an Anthropic employee quit over AI safety concerns and the issue went mainstream, Dr. Amodei preempted other companies and himself called on the industry to slow down.

About three weeks later, Dr. Amodei struck a different tone at the White House during a lunch with Mr. Trump and more than 30 other technology executives, venture capitalists and politicians.

Dr. Amodei, sitting near the end of a long table, said the United States must win in AI and that the technology should be built at maximum speed, according to two people familiar with the discussions.

It was, other guests thought, the opposite of what he was telling the world.

Cade Metz contributed reporting from San Francisco.