A group of 76 cybersecurity veterans published an open letter on freefable.org on June 15, urging the White House to revoke the export controls on Anthropic’s Fable and Mythos models. The letter states that the ban has “taken the best models out of the hands of defenders,” preventing them from using these tools to hunt for vulnerabilities and harden software and product security. It warns that “it is dangerous to strip defenders of their strongest capabilities without adequate justification while adversaries are rapidly advancing.” Signatories include former Facebook Chief Security Officer Alex Stamos, Bugcrowd founder Casey Ellis, renowned cryptographer Jon Callas, computer scientist Paul Vixie, former Block Application Security Engineering lead Dino Dai Zovi, Luta Security founder Katie Moussouris, and SocialProof Security CEO Rachel Tobac, among others.
At the heart of the controversy over the open letter is the evidence base for the ban: a non-public paper written by Amazon researchers that allegedly demonstrates a method for bypassing Fable’s safety guardrails. Moussouris pointed out in her blog that the paper actually demonstrates something far from a true jailbreak: after the model refused to identify security issues, the researchers only asked it to fix open-source code containing known, publicly documented vulnerabilities — “a find-fix-test loop that defenders perform every day, one of the most valuable things an AI model can do for cyber defense, not a guardrail bypass.” The open letter also cites analysis showing that the so-called bypass method works just as well on OpenAI’s GPT-5.5, Anthropic’s publicly available Claude Opus 4.8 and Sonnet, and even China’s Kimi 2.7 model, and calls on the government to establish export control rules based on scientific research through “democratic legislative processes,” limiting them to the minimum necessary to safeguard public safety.