100+ Security Experts Demand U.S. Reverse Anthropic AI Export Ban
One-line summary: More than 100 security professionals say the U.S. export ban on Anthropic's top models harms defenders more than adversaries — and that the "jailbreak" behind it was typing "fix this code."
Key Points
- Open letter organized by former Facebook CISO Alex Stamos, addressed to Commerce Secretary Howard Lutnick and National Cyber Director Sean Cairncross, gathered 100+ signatures from Nvidia, Google, and Adobe professionals
- Researcher Katie Moussouris, who reviewed the private paper behind the order, says the alleged vulnerability was simply a "fix this code" prompt — a standard defensive security workflow, not a novel attack vector
- Signatories argue the ban strips defenders of a high-value tool while doing nothing to stop adversaries, who have access to equivalent open-weight models
- Fable 5 has since returned in limited form with nationality-based access controls and tightened safety classifiers added
Why It Matters
If typing "fix this code" can meet the bar for a national security export control, the framework governing frontier AI models may be badly calibrated. The industry response here is likely to shape how the Commerce Department refines AI-specific export criteria — and how future model launches are structured to avoid overnight shutdowns.