The White House ordered Anthropic chief executive Dario Amodei to take down the company’s Fable model after a 19-day standoff over a jailbreak dispute, the first known case of the US government directly forcing a frontier AI laboratory to withdraw a deployed model.
Key points
- White House officials issued a direct order to Anthropic CEO Dario Amodei to withdraw the Fable model following a 19-day standoff over a jailbreak
- The request originated from Washington on the evening of April 6 and was relayed to Wall Street executives
- This represents the first known instance of the US government compelling a frontier AI lab to pull a deployed model
- The intervention marks a shift from voluntary industry safety commitments to executive enforcement
- Neither the White House nor Anthropic has publicly disclosed the specific jailbreak or the Fable model’s capabilities
The April 6 request and 19-day standoff
According to Politico’s detailed recap, an unusual request began moving from Washington to some of Wall Street’s most powerful executives on the evening of April 6. The request concerned a jailbreak involving Anthropic’s Fable model, though the specific nature of the jailbreak and the model’s capabilities have not been disclosed by either party. What followed was a 19-day standoff between US officials and the safety-focused laboratory founded by former OpenAI researchers.
From voluntary commitments to executive enforcement
Frontier AI laboratories have historically managed model safety and jailbreak responses through voluntary commitments and industry standards. Anthropic in particular has positioned itself as a safety-first laboratory, publishing research on constitutional AI and routinely submitting its models for external evaluation. The White House’s direct order to withdraw a deployed model represents a decisive shift from that voluntary framework to executive enforcement, establishing a new precedent for how the US government may intervene in frontier model deployment decisions.
Anthropic and the Fable model
Anthropic has not publicly acknowledged the Fable model’s existence prior to this reporting, nor has it detailed the jailbreak that prompted the government’s intervention. The company’s public communications have emphasised its responsible scaling policy and its practice of withholding models that do not meet safety thresholds. Whether Fable was a production model, a research release, or an internal system remains undisclosed. The White House has not issued a public statement on the order, and the legal instrument under which officials acted has not been identified.
The Politico report does not indicate a specific date for any follow-up action by the White House, Congress, or Anthropic regarding the Fable withdrawal or broader enforcement policy.
