Anthropic Refused to Patch Fable 5 Jailbreak Before US Export Ban

The Trump administration imposed export controls on Anthropic's most advanced AI models after CEO Dario Amodei refused to patch a jailbreak or pull the consumer‑facing product, according to a detailed account from David Sacks, co‑chair of the President's Council of Advisors on Science and Technology and the administration's former AI czar.

In a post on X on Saturday, as Tom's Hardware reported, Sacks said the administration issued the export control on Fable 5 and Mythos 5 "reluctantly" after Amodei declined to fix the flaw or de‑deploy the model. Sacks wrote that "the ball is in Anthropic's court" and the administration wants the restriction lifted once the jailbreak is patched.

According to Sacks, a trusted partner of both Anthropic and the US government — who was testing Fable — discovered a jailbreak that bypassed the guardrails separating the consumer‑facing Fable 5 from the unrestricted cyber capabilities of Mythos 5, the more powerful model it was built on. When the administration asked Amodei to fix the bypass or de‑deploy, he declined.

Sacks characterized the refusal as inconsistent with Anthropic's positioning as a safety‑first lab — the same company that had previously lobbied for Mythos to be regulated as a "cyberweapon." Anthropic defended its decision publicly, arguing the jailbreak is "narrow and non‑universal," and told Semafor that the same results could be produced on other public models, including OpenAI's GPT‑5.5.

Sacks rejected that argument outright: a bypass enabling operation of a cyberweapon is difficult to define as anything other than serious.

A person close to the White House told Semafor that Amazon flagged the jailbreak to the government, and that Amazon CEO Andy Jassy had been in contact with the administration about it. Amazon, which has invested billions in Anthropic and supplies much of its cloud computing infrastructure, did not confirm the details — a spokesperson told Semafor that governments often seek its counsel on security risks and that it does not discuss those conversations.

The White House acted partly over suspicion that a China‑linked group had accessed Mythos, raising the prospect of the model being reverse‑engineered or distilled. An An Anthropic spokesperson told Semafor the White House "didn't raise Chinese access to Mythos in its conversations around the Fable jailbreak" and that Anthropic blocks access to its products from inside China. This is not the first time Mythos access has leaked — unauthorized third parties reached the restricted model via a data breach in April.

The European Commission is now formally assessing the implications for EU users, with a spokesperson calling the episode a warning sign that Europe needs to strengthen its technological sovereignty.

"We are seeing a new generation of highly capable AI models reach the market. These models offer significant benefits, including for cyber‑defence, but they also raise serious cybersecurity concerns that need to be addressed," Thomas Regnier, European Commission spokesperson for tech sovereignty, told.² "We believe that contingency measures taken in this light should not be discriminatory against partners."

The Commission confirmed it is "looking closely at the practical consequences of this for European users of these services."

Anthropic is moving quickly to contain the political damage. Axios reported Sunday that the company has flown staff to Washington, D.C., to repair relations with the Trump administration. CEO Dario Amodei is scheduled to join G7 leaders and the chief executives of other leading AI companies for a working lunch on Tuesday, where the export controls are expected to be a central topic.

Canadian Prime Minister Mark Carney has already seized on the crisis, warning as The Next Web reports that the Anthropic ban demonstrates the risks of over‑reliance on a single AI provider, comparing it to the systemic linkages that triggered the 2008 financial crisis. Carney told AP News that countries must "build out and diversify" their AI supply chains.

For AI builders and developers outside the United States, the immediate impact is stark: access to Anthropic's most capable models is now blocked. If you were building on Claude Fable 5 or Mythos 5 APIs from Europe, Canada, India, or anywhere outside the US, those endpoints are dark.

Sacks has indicated, Tom's Hardware reports, that the administration wants the controls lifted quickly — "the ball is in Anthropic's court" — but the timeline depends on whether Anthropic patches the jailbreak to the government's satisfaction. In the meantime, European Commission officials are assessing the precedent this sets. The episode also pulls forward long‑simmering questions about whether AI models should be subject to export controls at all, and what a fragmented global AI landscape looks like for the developers who build on top of these models.