Anthropic Allegedly Sparked U.S. AI Export Ban

NEW YORK — The AI export ban on foreign users of Anthropic’s newest models has sparked a fresh debate in Washington, after the company was seen as campaigning too aggressively about the risks of artificial intelligence. The Financial Times reported that criticism of Anthropic has grown because the warnings it raised may have helped lead to restrictions on access to the Mythos and Fable models.

In an analysis of official statements, social media posts, and writing from Anthropic and its chief executive, Dario Amodei, the FT found a sharp contrast with OpenAI. Throughout 2026, every 1,000 words of Anthropic’s communications contained about five words related to risk, regulation, or restrictions. OpenAI and Sam Altman used only 0.6 such words per 1,000 words.

Anthropic’s cautious language has started to backfire

The number sounds small. But in tech-company messaging, word choice can carry real weight. Anthropic has long positioned itself as the industry’s most cautious voice on AI, frequently raising concerns about danger, oversight, and the need for government intervention. That approach built a reputation for seriousness on safety. Now it is being used by opponents to argue that Anthropic helped open the door to tighter controls.

Tensions rose after Washington last week barred foreign nationals from using Anthropic’s newest models, Mythos and Fable. Some technologists say the move was directly tied to Anthropic’s repeated warnings about AI risks, especially around Mythos. Yann LeCun, Meta’s former chief AI scientist, even called Dario Amodei’s warnings “fear-mongering” and added, “One reaps what one sows.”

That kind of comment is more than an academic spat. For industry players, it can become a signal to regulators, investors, and government partners. Once that signal reads as a security threat, access can close fast. Very fast.

FT data shows Anthropic talks about risk more often

The FT compiled words such as “harmful,” “dangerous,” and “misaligned,” then counted how often they appeared in the two companies’ public communications. The result: the word “risk” appeared 336 times in Anthropic’s 2026 communications. “Safeguard” appeared 121 times, while “vulnerability” appeared 128 times. At OpenAI, those three words appeared only 30 times, 33 times, and 10 times.

The FT also ran a sentiment analysis to compare positive and negative tone. The overall picture showed Anthropic’s public communications were still mostly positive, but more negative than OpenAI’s. Interestingly, Anthropic’s language has softened since 2023, as its AI tools became more popular. The frequency of words tied to risk and regulation fell by about half from the same period three years earlier.

That means Anthropic does not always speak in the darkest terms. Yet in the debate over Mythos and Fable, its record of warnings still sticks. Once a reputation forms, it is hard to scrub away. Especially after policy has already turned into an actual ban.

Mythos, Fable, and the political pull in Washington

Anthropic has promoted Mythos as a model capable of finding important cybersecurity flaws. At first, access was limited to selected organizations in the U.S. on safety grounds. The company also reportedly worked with government officials on a supervised launch before releasing Mythos more broadly earlier this month.

The process was not smooth. Some industry voices criticized how Anthropic negotiated with the government over its new model. David Sacks, a former AI adviser to the U.S. government, wrote on X that a “trusted partner” had approached the administration seeking ways to get around Fable’s safety guardrails. He accused Anthropic of downplaying those concerns, saying the government was “forced” to impose a ban.

From there, a technical issue turned political. The ban does not only target a product. It also touches the larger direction of U.S. AI policy: how open frontier models should be to the world, who gets to use them, and when security should outweigh innovation.

The U.S., Europe, and concern over frontier model access

Washington’s decision has alarmed some in Europe and Silicon Valley. They fear the Trump administration may be willing to restrict non-U.S. access to the most advanced AI models. For executives and officials, this could be an early test of how the U.S. oversees AI models that are more powerful, more expensive, and more strategically important.

A YouGov survey cited by the FT found that a majority of respondents agreed effective regulation matters, even if it slows technology. At this point, the public does not automatically reject restrictions. The real issue is the method and consistency. If AI companies themselves promote a danger narrative, then the government shuts access, who is actually in control?

French President Emmanuel Macron has also entered the debate. He said the Anthropic dispute had “made the stakes clear” for the U.S. and its G7 allies. Macron called for “stronger regulation of artificial intelligence” and warned about the risk of “non-cooperation among democracies.” His message was blunt: without coordination, democratic countries can end up mistrusting one another as AI advances too quickly.

Lennart Heim, an independent AI policy researcher who previously worked at Rand, said the U.S. government’s response did “not inspire confidence.” In his view, a government that has cast itself as pro-innovation, pushed for exporting advanced AI chips to China, and criticized safety-based regulation has now reversed course by banning the most advanced U.S. models. That contradiction, he said, is hard to explain to the public.

What this means for the AI industry ahead

The Anthropic case shows one simple thing: in AI, a company’s own words can flip into state policy. When a company stresses danger too often, regulators may see that as a reason to act. When it downplays risk, the public accuses it of recklessness. Either way, it is a trap. And it is in that space that the industry’s direction gets set.

For readers, the impact is concrete. Restricting access to frontier AI models is not just a matter for giant firms in San Francisco or officials in Washington. Decisions like this can shape who gets to test the latest models, how outside researchers work, and how quickly AI features reach everyday products.

Anthropic now sits in a difficult spot: it wants to be seen as an AI ethics watchdog, but it also does not want to be blamed for an AI export ban. Going forward, observers will be watching not only Mythos or Fable performance, but also whether the U.S. government starts treating model access restrictions as a standard tool. That will be the next big story.

Brief summary: Anthropic talks about AI risk more often than OpenAI, according to Financial Times analysis. Those warnings are intertwined with the U.S. government’s ban on foreign access to the new Mythos and Fable models. The case is now an early test of how Washington regulates the most powerful AI systems.

Short FAQ: Why does this matter? Because model-access decisions can affect innovation, safety, and tech diplomacy. What should be watched next? The U.S. government’s stance on other frontier models and whether similar restrictions spread.