Feds Tell Anthropic: Fix Fable 5's Guardrails or Forget It
Trump administration demands Anthropic prove Fable 5's safety controls are bulletproof before the model can return.
The Trump administration has drawn a hard line with Anthropic. Officials told Wired that if the AI company wants to rerelease its Fable 5 model, it must first guarantee the model's guardrails cannot be circumvented.
That's a tall order. Experts say achieving truly unbreakable guardrails may simply not be possible. AI safety researchers have long demonstrated that determined users can find ways around model restrictions through jailbreaks and prompt injection techniques.
The standoff puts Anthropic in a difficult position. The company needs government approval to bring Fable 5 back, but the technical bar being set may be fundamentally unachievable with current methods.
It's a classic regulatory catch-22: prove something works perfectly in a domain where perfection doesn't exist. No timeline has been set for resolution.