White House Demands Anthropic Block All AI Jailbreaks, But Experts Say It May Be Impossible
The Trump administration is pushing Anthropic to fix vulnerabilities in its Claude Fable 5 AI model, but security experts argue that completely preventing jailbreaks may not be feasible.
Dispute Between Trump Administration and Anthropic Escalates
Trump administration officials have stated that if Anthropic wants to re-release its AI model Claude Fable 5, the company must take steps to address alleged vulnerabilities. The model was taken offline last week under export controls due to concerns about jailbreaking—using prompts to bypass safeguards.
Anthropic has maintained for days that the administration's concerns are overblown and the effects of jailbreaks are minimal. The company reiterated this stance during a technical meeting Monday with the Commerce Department and the Office of the National Cyber Director.
However, officials say the debate over whether the jailbreaks are significant is over, as the National Security Agency (NSA) has concluded there are ways to disable guardrails on Fable 5. These guardrails are meant to prevent access to capabilities of the Mythos model related to cybersecurity, chemistry, and biology.
Administration's Demands and Expert Opinion
According to three people familiar with the discussions, the administration views the situation as Anthropic's problem to fix. Neither the Commerce Department's Center for AI Standards and Innovation nor the NSA has the staff or bandwidth to chase down every conceivable jailbreak on every model that reaches the market.
Consequently, the administration believes Anthropic should be more proactive in continuously testing not just Fable 5 but all its frontier AI models to find potential jailbreaks and report them to the government.
On a more fundamental level, it remains unclear how Anthropic is supposed to prevent jailbreaking. Independent cybersecurity experts increasingly see guardrails as a stopgap solution, as skilled users and future AI models will find ways to bypass constraints. This suggests that what the White House wants may not be achievable.
A White House spokesperson declined to comment.
