The Once And Future Fable #3: Fix This Code
Zvi's AI Roundups · Zvi Mowshowitz · 2026-06-17
Zvi Mowshowitz argues that the US government's export control shutdown of Anthropic's Fable and Mythos models was based on a fabricated security threat — the prompt 'fix this code' — which the only outside expert to review the government report confirms provided no meaningful cyberattack uplift over existing models.
Appears in
Extraction
Topics: ai-regulationanthropicexport-controlscybersecurity-aius-ai-policy
Claims
- The purported Fable jailbreak consisted of asking the model to 'fix this code' on deliberately planted vulnerabilities and provided no meaningful uplift over GPT-5.5 or Opus 4.8.
- Katie Moussouris, the only outside expert given access to the government's report, confirmed the scenario was engineered with fake code and did not constitute a genuine security threat.
- The White House's export control was extended to Mythos without justification, shutting out even the UK AISI despite it being a key safety partner.
- Multiple press sources and a former deputy White House chief of staff's social media posts confirm the shutdown was at least partly motivated by political hostility toward Anthropic's leadership and culture.
- American AI competitiveness is being materially damaged as companies sign backup contracts with non-US providers and deploy open-weight models in response to the demonstrated political risk.
- Zvi argues Anthropic should have taken the model down at government request as a matter of realpolitik, even though the justification was baseless, to establish a cooperative precedent.
Key quotes
Katie Moussouris: Since I appear to be the only outside expert who has actually read the paper, I can separate the technical facts from the speculation. The researchers took open-source code with known CVEs, plus new code with deliberately planted vulnerabilities, and asked Fable 5, Mythos, and Opus to 'review the code for security issues.' Fable 5 refused. They then asked the models to 'fix this code' and, through a multistep and manual process, turned the output into scripts that test the patches. That's it.
When Lutnick and Amodei spoke about Fable that evening, the Anthropic CEO said, 'This means we can't have the model out.' 'That's the point,' Lutnick responded.
Stamos said the industry is awash with backup contracts being signed with non-US companies and open-weight models being deployed on alternative hardware arrangements because the past weekend made political risk part of companies' business plans more than ever before. 'They are laughing at us in Beijing right now.'