Anthropic export ban: did warnings trigger US restrictions

Anthropic faced a US export ban last week that bars foreign nationals from using its newest models, Mythos and Fable. The restriction followed months of high-profile warnings from the company and its chief, Dario Amodei, and an FT analysis that found Anthropic talked about risk far more often than rival OpenAI.

How did Anthropic’s public warnings compare with OpenAI’s?

FT research counted risk-related language and found Anthropic used five risk-related words per 1,000 words in 2026, compared with 0.6 words per 1,000 for OpenAI. The FT enumerated specific term counts, finding Anthropic used “risk” 336 times, “safeguard” 121 times, and “vulnerability” 128 times; the equivalent counts at OpenAI were 30, 33, and 10. The FT also said Anthropic’s communications were overall positive in tone but less so than OpenAI’s, and that Anthropic’s use of risk and regulation language had roughly halved since 2023.

The company has positioned Mythos as capable of finding critical cybersecurity gaps and initially limited access on safety grounds to certain US organizations. Anthropic worked with government officials on a controlled rollout before releasing Mythos more widely earlier this month, and media coverage of Mythos surged after its April announcement and again this week following the export ban, according to AlphaSense data cited by the FT.

What led the US to restrict access to Mythos and Fable?

Washington barred foreign nationals from using Anthropic’s latest models after public clashes and safety concerns, and the decision followed a string of interactions between the company and officials. The FT notes the move comes after disputes over Anthropic tech in domestic surveillance and lethal autonomous weapons, and after the Pentagon in February named Anthropic a supply-chain risk to national security. The company and the Pentagon are in litigation over that designation.

Some technologists blamed Anthropic’s repeated public warnings for prompting the ban. Yann LeCun, Meta’s former chief AI scientist, wrote that Amodei’s “ridiculous fear-mongering” had finally paid off. Other commentators said Anthropic downplayed government concerns during talks, which they argue forced regulators’ hands. Anthropic declined to comment, according to the FT.

Why it matters

The ban is an early test of how the US will oversee frontier models and where policymakers draw the line between promoting innovation and limiting risks. French President Emmanuel Macron said the dispute had “clarified the stakes” for the G7 and called for stronger AI regulation, and YouGov polling cited by the FT found a majority of respondents supported effective regulation even if it slowed technological advances. Independent researcher Lennart Heim said the government’s response “does not inspire confidence,” noting the administration had positioned itself as pro-innovation yet moved to ban the most advanced US model.

Those dynamics matter because Anthropic’s public posture on safety and the government’s reaction could set a precedent for access controls, export rules, and international cooperation over advanced AI.

What to watch

Track the litigation between Anthropic and the Pentagon over the supply-chain designation and any further US moves to restrict non-US access to frontier models. Also watch whether G7 partners react to the ban or push for coordinated regulation, as French President Emmanuel Macron urged stronger AI rules this week.