Anthropic warns: Fighting with Trump administration could hit revenue

AI giant Anthropic looks nervously towards Washington just before its IPO. Chatbot creator Claude warns this week that his open advocacy for secure AI could make him a target of the U.S. government.

The money Anthropic earns from the government is small. The risk clearly is not.

The warning is in the prospectus that Anthropic filed on Monday. This is the mandatory document in which a company tells investors about opportunities and risks.

According to Anthropic, his views on AI surveillance and the statements of his top executives can put considerable pressure on the government. The company writes the following

“We have faced, and may continue to face, increased visibility and may continue to be subject to heightened scrutiny or the risk of being disproportionately or selectively addressed by certain government agencies, even where similar companies are not.”

The fight started earlier this year at the Pentagon. Anthropic refused to use its AI for autonomous weapons and mass surveillance.

In response, Defense Secretary Pete Hegseth declared the company a supply chain risk in March. Anthropic went to court and, according to the AP news agency, spoke of an "unlawful retaliatory campaign".

That battle is still ongoing. A judge in California ruled Anthropic right, but an appeals court in Washington allowed the Pentagon to take Claude out of its systems. Defence suppliers must also follow suit.

Then came Mythos, an AI model that finds software leaks so good that Anthropic only shared it with a handful of partners. After disagreements over security, the government imposed export restrictions in June.

Anthropic then had to temporarily shut down Mythos 5 and Fable 5 completely. After more than two weeks, everything was allowed to open again, but according to the company, this could take longer next time.

On paper, the damage is not too bad. Government contracts account for about 1 percent of sales: out of the 16.3 billion dollars from the first half of the year, this is about 160 million dollars.

Still, that number doesn't tell the whole story. Companies that supply the military also have to dump Claude, and Anthropic does not rule out that the lawsuits about that will affect its business sale.

In addition, CEO Dario Amodei argues for a joint brake on AI. Donald Trump and his allies call the fear that AI threatens humanity a “hoax,” while Anthropic names those risks in its own prospectus. The president mainly wants to win the race with China.

Last month, Amodei and more than twenty other tech executives in the White House had to sign a voluntary security agreement. Days later, the government threatened to take steps against companies that did not adhere to it.

Sarah Kreps, director of Cornell University's Tech Policy Institute, sees a problem in this: “With AI, the boundaries of government authority are still being explored, sometimes through direct confrontations with individual companies. It's hard for investors to put that kind of uncertainty into the price.”

Meanwhile, the market is counting on a valuation of around 2,000 billion dollars, more than double the 965 billion from the last financing round.

Crypto traders are already close to that. On platforms like Binance and Hyperliquid, they price the company at nearly $1.9 trillion, with contracts that track the stock's future price.

The debut on the Nasdaq, under the ticker ant, is scheduled for the week of November 16.