The interview had barely aired when the phones started ringing at Anthropic’s offices. President Trump told Axios he no longer regards the AI company as a national security threat, a reversal of the position that had placed the firm on a government watchlist and cast a shadow over its international expansion. Company executives welcomed the shift in public statements, but inside the firm, according to people familiar with the matter, the mood was more complicated: the same week, the Commerce Department ordered Anthropic to stop offering its newest flagship models abroad.
The models in question are Fable 5, the company’s most advanced large language model, and Mythos 5, a related release. Forbes confirmed that both have been taken offline in international markets after the Commerce Department instructed Anthropic to halt their overseas availability. The company has complied, a person close to the firm said, while disputing the factual basis for the order.
The sequence is hard to read as a single policy. On one track, the administration is courting Anthropic, whose co-founders have argued for years that AI development should continue under careful oversight rather than pause, a position that aligns with the White House’s preference for American leadership in the technology. On the other track, parts of the government are treating the company’s most capable models as export-controlled commodities, restricted from foreign users in the name of national security.
Politico, citing legal analysts, reported that the restrictions may run afoul of the law. The export-control framework governing AI models is built around thresholds and review processes, and applying it to specific commercial products on short notice, without published criteria or an appeals path, raises due-process questions that courts have begun to examine in related cases. Legal scholars quoted in the report said the Commerce Department’s action against a company the White House is simultaneously embracing is difficult to square with either the letter of the regulations or the administration’s public posture.
For Anthropic, the practical stakes are substantial. International markets account for a growing share of its revenue, and Fable 5 was positioned as the company’s answer to rival frontier models on global app stores and cloud platforms. Pulling it from those markets hands ground to OpenAI, Google and a handful of Chinese labs whose models face fewer restrictions in most countries. The company has told employees it is working with the Commerce Department to resolve the matter, according to people familiar with the internal communications.
The political context has shifted faster than the regulatory machinery. Anthropic spent much of the past year under a cloud, reportedly reviewed by national security agencies over concerns about foreign investment and the dual-use nature of its technology. The Axios interview effectively closed that chapter, with the president describing the company in favorable terms and signaling that the White House wants it as a partner rather than a target.
What follows is a negotiation over the rules themselves. Administration officials have said they want to work with Anthropic on AI safety standards, according to people familiar with the discussions, and the company has been an active participant in talks about how frontier models should be tested, monitored and, where necessary, restricted. The Fable 5 episode gives both sides a concrete case study for where those rules break down.
The tension inside Anthropic is real. The company’s founding mission is to build AI that is safe, and it has historically supported the idea that capable systems warrant scrutiny. But being the subject of that scrutiny, in a way that removes its best products from paying customers overseas, tests the company’s patience with the very framework it helped popularize. Employees have debated the trade-offs openly, according to people familiar with the internal channels, with some arguing the company should fight the order publicly and others urging quiet resolution.
The outcome matters beyond one company. If the White House and the Commerce Department settle on a negotiated framework with Anthropic, it could become the template for how American frontier models are governed internationally — an industry-wide standard set through a single company’s case. If the dispute festers, it could chill every AI lab’s plans to sell its best models abroad, and hand the global market to competitors with fewer scruples about who uses their technology.
The company’s path to the White House has been anything but smooth. Anthropic was reported last year to be under national security review, a designation that complicated its dealings with foreign partners and forced its executives to spend months reassuring officials about the company’s governance and the sources of its capital. The president’s Axios remarks were the first public signal that the review had concluded, and administration officials have since described Anthropic’s leadership as cooperative. The timing of the Commerce Department’s order, landing in the same week as the political rehabilitation, has left even sympathetic observers wondering which branch of the government speaks for the United States on AI.
Neither the White House nor the Commerce Department responded to requests for comment. Anthropic said it remains committed to working constructively with the administration on safety standards while it resolves the questions around Fable 5 and Mythos 5. The models sit offline for now, and with them a test of whether the government’s two hands are speaking to each other.
