Tech news in 3 minutes
Cybersecurity vets protest ‘dangerous’ US government ban on Anthropic’s most powerful models
A group of 76 cybersecurity experts, including industry veterans like former Facebook chief of security Alex Stamos and Bugcrowd founder Casey Ellis, has published an open letter urging the U.S. government to lift an export control order on Anthropic’s Fable and Mythos AI models. The letter argues that the order, issued last Friday on national security grounds without specific explanation, harms cybersecurity defenders by removing the best tools for finding vulnerabilities and securing software. “To pull the best capabilities away from defenders without a good reason when our adversaries are rapidly advancing is dangerous,” the letter states. In response to the order, Anthropic suspended access to both models worldwide. The experts claim the export ban is based on a non-public Amazon paper that purportedly demonstrates a method to bypass Fable’s guardrails and unlock Mythos-level capabilities. However, Katie Moussouris, a signatory who reviewed the paper, asserts it does not show a real jailbreak; instead, researchers asked Fable to fix known vulnerabilities after the model refused to review code for security issues. She argues that such behavior is essential for defensive security and cannot be fixed without weakening the model. The letter also notes that similar capabilities can be replicated on other models like OpenAI’s GPT-5.5, Anthropic’s Claude Opus 4.8, or China’s Kimi 2.7. The open letter calls for transparent, fairly enforced regulations developed through a democratic process based on scientific research, used only to the minimal extent necessary for public safety. Anthropic previously restricted Mythos to about 50 companies, later expanding to 150 organizations in 15 countries. Fable, a public version with strict guardrails, was released last week but has been criticized for blocking nearly all cybersecurity-related prompts.