The Anthropic export control ban arrived at 5:21 PM ET on 12 June, when the Commerce Department sent a letter to CEO Dario Amodei invoking national security authorities to bar all foreign nationals from accessing Fable 5 and Mythos 5, the company’s two most advanced AI models. The directive applied not just to users abroad but to foreign nationals inside the United States, including Anthropic’s own foreign-born employees. Faced with no practical way to screen every user, Anthropic pulled the models entirely.
The immediate trigger, according to multiple reports, was Amazon. Reuters reported that Amazon CEO Andy Jassy was among tech leaders who raised concerns with senior Trump administration officials about security risks in Anthropic’s models, citing a person familiar with the matter. More specifically, the Wall Street Journal, as reported via mlq.ai, identified Jassy’s contact as Treasury Secretary Scott Bessent, not the White House directly: Amazon cybersecurity researchers had reportedly jailbroken Fable 5 to extract information usable in cyberattacks, and Jassy passed that finding to Bessent. Things spiralled quickly from there, on a Friday afternoon, into a weekend.
Fortune also reported that concerns about unauthorised Chinese access to Mythos were among the factors the administration cited, per Semafor reporting. The government’s stated reasoning remains vague: the letter cited national security concerns, gave no specifics, and has not been made public.
Why the Anthropic Export Control Ban Looks Like More Than a Security Call
This ban did not materialise from a clear sky. On 27 February 2026, President Trump directed federal agencies to stop using Anthropic technology, and Secretary of Defense Pete Hegseth designated Anthropic a ‘Supply-Chain Risk to National Security,’ following a prolonged dispute over the company’s refusal to let its AI be used for certain military applications. The Congressional Research Service documented that standoff, and The Verge reports that the friction centred on Anthropic’s refusal to allow its AI to power mass surveillance of Americans or lethal autonomous weapons.
That backstory matters. AI policy expert Dean Ball, who briefly served in the Trump administration, told Fortune he could not determine whether the export control move amounted to ‘lawfare against Anthropic in particular or extreme national-security hawkery,’ and called it ‘simply cartoonish.’ That is a polite way of saying: when the legal mechanism chosen to address a jailbreak also bars a company’s own foreign-born staff from touching their own product, the proportionality argument is hard to make with a straight face.
Anthropic itself noted that the same jailbreaking methods had been found in several other AI models, a point that cybersecurity researchers echoed loudly. 76 cybersecurity experts signed an open letter urging the government to revoke the order, arguing it was actively dangerous to strip advanced cybersecurity capabilities from network defenders inside the United States, according to mallory.ai. Open-letter signatory Katie Moussouris told TechCrunch that the Amazon jailbreak method was demonstrated in a research paper she had personally reviewed, but which is not yet public.
Who Actually Benefits Here?
On paper, Anthropic’s competitors should be pleased. If the administration’s appetite to invoke export controls is tied to a specific, fractious relationship with one company rather than a consistent policy applied across the sector, then OpenAI, Google DeepMind, and others can breathe a little easier. ‘Anthropic has not had the best relationship with the Trump administration in a way that stands apart from the other leading AI labs,’ as Sean O’Kane put it on the Equity podcast. The implied corollary: stay friendly, stay operational.
That is cold comfort for the industry as a whole. Regulatory exposure that depends on personal chemistry between a CEO and a cabinet secretary is not a framework, it is a risk. Companies that think they have avoided this simply by not annoying the right official are operating on borrowed goodwill.
And then there is the stranger possibility: that Anthropic itself benefits. Fable 5 was built as a public-facing version of Mythos with strict guardrails blocking use in biology, chemistry, and cybersecurity, and preventing model distillation by third parties. The government effectively told the world that a company’s most locked-down, guardrail-heavy product was still too dangerous to exist. The previous flashpoint between Anthropic and the Trump administration, the podcast team noted, sent Claude downloads surging. Being labelled radioactive by a White House that has its own complicated relationship with tech credibility is not the worst marketing outcome. As Rebecca Bellan put it: ‘Everyone loves a bad boy.’
The open letter gives the administration an off-ramp. Whether it takes it, or doubles down, will tell you rather a lot about whether this was really about security at all.
