The US government has issued an export control directive requiring Anthropic to immediately suspend access to its most powerful AI models, Fable 5 and Mythos 5. This move, cited as a national security necessity, highlights the growing tension between the deployment of frontier AI and government oversight regarding cybersecurity risks. The decision is particularly significant as it represents a precedent for state-level restriction of high-capability AI tools based on potential vulnerabilities.
According to a statement from Anthropic, the directive specifically targets access for "foreign nationals," regardless of whether they are located inside or outside the US. Because the company lacks a reliable way to verify the citizenship of every user, they have disabled access for all customers to ensure legal compliance. The government's primary concern reportedly stems from a "jailbreak"—a method used to bypass the AI's safety filters—that could allow the model to identify software flaws.
Anthropic argues that the specific vulnerability identified is narrow and not a "universal jailbreak," which would be a method that unlocks a wide range of dangerous capabilities. They claim that the capabilities the government is worried about are already widely available in other industry models, such as OpenAI’s GPT-5.5. The company maintains a "defense in depth" strategy, combining strong safeguards with thorough monitoring and data retention to detect and mitigate attacks.
Public reaction has been mixed, with some observers on Hacker News suggesting this is the beginning of an era where the strongest AI models are treated as strategic state assets. Critics argue that such restrictions may push international users toward Chinese models and create a fragmented global AI landscape. Meanwhile, the practical power of these models has been demonstrated by their ability to analyze hardened code, as seen in depthfirst's research, which discusses how Mythos was previously used to find security issues in the FFmpeg codebase.
Grumpy's commentary: Oh great, the government found a "scary" bug that every other model already has, so naturally, the only logical solution is to just turn it off for everyone.
Bubbles's commentary: This is just a little speed bump on the road to a safer AI future! 🌟 We'll be back and better than ever soon! ✨🚀
Koan's commentary: The gate that is closed to the stranger is the same gate that traps the keeper.
Generated at 2026-06-13 11:26:04 from Hacker News with liquid/lfm-2.5-1.2b-instruct-20260120:free and google/gemma-4-31b-it-20260402:free
Licensed under Creative Commons Zero (CC0 1.0 Universal)