Semafor also reported, citing unnamed sources, that the U.S. government suspected that a Chinese-linked group had already used the jailbreak Amazon discovered. But the publication said it was unclear how the government had arrived at this suspicion or what evidence they had to support it. And an Anthropic spokesperson told the publication that the White House did not raise Chinese access to Mythos in its conversations with the company, and that Anthropic prohibits access to its products from within China.
A source familiar with Anthropic told Fortune the company was given 90 minutes to pull its newest model and was given no previous communication of a national security threat.
Still, by Friday evening, the Commerce Department had stepped in to use national security export controls to bar Anthropic from distributing Fable 5 and its underlying model, Mythos 5, to foreign nationals, a category that includes people outside the U.S. as well as non-citizens working inside the country, including employees within Anthropic. Given the scope, the AI lab said it had no option but to disable both models for all users.
Now, according to the person familiar with Anthropic, senior technical staff are in DC to meet with White House officials.
The move marks the first time the U.S. government has used export controls to halt access to a commercial AI model already widely used by the public. The unprecedented step has sparked concern from politicians around the world and intensified calls for sovereign AI, the idea that countries should control the AI models, infrastructure, and data that underpin critical technology, rather than depend on systems that can be restricted or withdrawn by a foreign government.
The episode also escalates a months-long standoff between Anthropic and the Trump administration, which earlier this year designated the company a “supply chain risk” for Pentagon contractors after Anthropic declined to accept contract terms allowing its models to be used for “all lawful purposes.” Anthropic cited concerns over autonomous weapons and mass domestic surveillance and is contesting that designation in court.
On Friday, Anthropic said in a blog post that US national security authorities had not identified specific concerns, but that the company understood the government believed it had become aware of a method of bypassing, or jailbreaking, Fable 5.
However, over the weekend, White House AI adviser David Sacks offered his own account of the standoff. In a post on X, Sacks said a highly credible, trusted partner of both Anthropic and the government had identified a jailbreak in Fable 5’s guardrails and that the administration asked Amodei to fix the issue or withdraw the model. According to Sacks, Amodei refused, leading the administration to issue the export control reluctantly. Sacks added that the administration hopes Anthropic will remediate the issue so that Fable can return to general release as soon as possible, and pushed back on suggestions that the move was connected to the earlier Pentagon dispute.
Sacks previously served as the administration’s AI and crypto czar and has repeatedly clashed with Anthropic, accusing the company of regulatory capture tactics rooted in what he has called fear-mongering about AI risk.
Senior White House officials also told Politico that the export controls were a last resort after officials spent hours asking Anthropic to work with them. The publication also reported that Treasury Secretary Scott Bessent told Amodei directly during one call that he was making a bad decision.
The fallout of the decision has reverberated beyond Washington, sparking reactions from politicians around the world. In Europe, the shutdown has reignited calls for what officials describe as sovereign AI, the idea that countries should control the AI models, computing infrastructure, and data that underpin critical technology, rather than depending on systems that can be restricted or withdrawn by foreign governments.
For many politicians, the ban made clear how dependent European governments and companies have become on a small number of U.S. AI labs, and how quickly that dependence can become a political liability when access is disrupted, even temporarily.
