A coalition of prominent cybersecurity experts has issued an open letter to the U.S. government, urging the reversal of an export control order placed on Anthropic’s advanced AI models, Fable and Mythos. The experts contend that this directive, citing national security concerns, effectively disarms digital defenders by removing crucial tools necessary for identifying vulnerabilities and enhancing software security.

The U.S. government’s order, issued recently, prompted Anthropic to suspend global access to both Fable and Mythos. While the specific reasons for the national security concerns remain undisclosed by the government, Anthropic suggested the decision might stem from reports of a method to bypass, or ‘jailbreak,’ Fable to unlock its more powerful Mythos-level capabilities. Mythos, initially launched with restricted access to about 50 companies due to its potent vulnerability-finding abilities, was designed to prevent misuse by malicious actors. Fable, a public iteration, included strict guardrails to limit its application in sensitive fields like cybersecurity, which many experts found overly restrictive.

However, signatories of the open letter, including former Facebook chief of security Alex Stamos and cryptographer Jon Callas, challenge the premise of a ‘jailbreak.’ Katie Moussouris, founder of Luta Security and a key signatory, reviewed a non-public paper by Amazon researchers that allegedly demonstrated such a bypass. Moussouris argues that the paper merely showed Fable fixing open-source code with known vulnerabilities, a standard defensive task, rather than a true security bypass. She emphasized that enabling AI to find, fix, and test bugs is a vital function for defensive security, not a guardrail violation. The experts further assert that similar capabilities are present in other widely available models, including OpenAI’s GPT-5.5, Anthropic’s own Claude Opus 4.8 and Sonnet, and even Chinese models like Kimi 2.7.

The letter concludes by advocating for transparent, democratically established regulations based on scientific research from industry and academic experts. These regulations, the experts argue, should be minimally intrusive and solely focused on ensuring public safety, rather than inadvertently hindering the very defenders they aim to protect.

Key Takeaways

• The U.S. government has banned the export of Anthropic's Fable and Mythos AI models, citing national security concerns, leading to their global suspension.
• Dozens of cybersecurity experts are protesting the ban, arguing it removes essential tools from defenders needed to find and fix software vulnerabilities.
• Experts, including Katie Moussouris, dispute the government's potential 'jailbreak' rationale, claiming the demonstrated method is a legitimate defensive use case and similar AI capabilities are available in other models.

Editor’s Analysis & Impact

This situation highlights a critical tension between national security imperatives and the rapid advancement of AI technology, particularly in the cybersecurity domain. The government’s export ban on Anthropic’s models, while intended to prevent misuse, risks inadvertently weakening the defensive capabilities of organizations that rely on such tools to combat increasingly sophisticated cyber threats. The experts’ argument that similar capabilities exist in other models, coupled with their critique of the ‘jailbreak’ interpretation, suggests a potential disconnect between policy-making and the practical realities of AI application in security. This event could set a precedent for future AI regulation, emphasizing the urgent need for clear, scientifically informed policies that balance innovation, national security, and the operational needs of cybersecurity professionals.

Frequently Asked Questions

Q: What are Anthropic's Fable and Mythos AI models?
A: Fable and Mythos are advanced artificial intelligence models developed by Anthropic, designed to identify and help fix security vulnerabilities in software and systems. Mythos was the more powerful, restricted version, while Fable was a public release with built-in guardrails.

Q: Why did the U.S. government ban the export of these models?
A: The U.S. government issued an export control order on Fable and Mythos citing national security concerns. While specific reasons were not detailed, Anthropic suggested it might be related to reports of a method to 'jailbreak' Fable to access its more potent capabilities.

Q: What is the main argument of the cybersecurity experts against the ban?
A: The cybersecurity experts argue that the ban removes critical tools from digital defenders, hindering their ability to secure software and products. They also contend that the alleged 'jailbreak' method is a legitimate defensive use case for AI and that similar vulnerability-finding capabilities are already present in other widely available AI models.