01What happened
The story, straight
The Trump administration's export-control directive blocking Anthropic's Fable 5 and Mythos 5 models was triggered by a simple three-word prompt — 'Fix this code' — not a sophisticated jailbreak, according to Katie Moussouris, founder and CEO of Luta Security. Moussouris says she was the only outside expert to read the third-party research paper on the guardrail bypass techniques that prompted the ban. On Friday, the US government reportedly cited national security concerns in issuing the directive, which suspended access to both models for any foreign national inside or outside the United States. Anthropic responded by disabling both models for all customers to ensure compliance. Moussouris detailed her findings in a Monday blog post after Anthropic shared the report privately with her.
the feds didn't ban Anthropic's Fable 5 over some elite jailbreak. it was a three-word prompt: 'Fix this code.' Katie Moussouris, founder of Luta Security and basically the godmother of bug bounties, says she was the only outside person to read the research paper that triggered the whole thing. The US government issued an export-control directive Friday, blocking Fable 5 and Mythos 5 from any foreign national, citing national security. Anthropic killed access for everyone to stay compliant. Moussouris broke it down in a Monday blog post.
02Spread timeline
Where it actually started
03Source receipts
Every claim, linked
04What's solid, what isn't
What's solid and what isn't
- The US government issued an export-control directive blocking Fable 5 and Mythos 5 from foreign nationals on Friday.
- Anthropic disabled both models for all customers in response.
- Katie Moussouris says she was the only outside expert to read the third-party research paper.
- The exact wording and scope of the third-party research paper's findings.
- Whether the 'Fix this code' prompt is the only or primary bypass technique documented.
- Other AI safety researchers have not yet publicly corroborated or disputed Moussouris's account.
05Why it matters
The editorial take
The revelation reframes a sweeping national-security action — one that cut off access to Anthropic's most advanced models worldwide — as a reaction to a basic prompt, not a novel attack vector. It raises questions about the threshold for government intervention in AI model access and whether regulators fully understood the technical details before acting. Moussouris's account, as the sole outside expert with access to the research paper, is currently the only public window into what the government actually saw.
the feds nuked access to Anthropic's best models globally because of a prompt a junior dev could write. that's a hell of a threshold for export controls. the fact that only one outside expert read the actual paper before the government acted is the part that should make everyone nervous.