US Cybersecurity Experts Warn Against Ban on Anthropic AI Models Mythos 5 and Fable 5

The US government has banned the use of Anthropic’s Mythos 5 and Fable 5 AI models over cybersecurity concerns. Anthropic has suspended these models for all users. The White House fears that malicious actors could jailbreak these models and use them to hack software systems.

Key Highlights

• US bans Anthropic’s Mythos 5 and Fable 5 AI models over cybersecurity concerns.
• 160 cybersecurity experts argue the ban could harm US and allied cyber defenders.
• Experts warn Chinese AI models may already rival or surpass US capabilities.
• AI models like GPT-5.5 and Kimi 2.7 can also find cybersecurity flaws.
• Anthropic is in talks with US officials about the potential reversal of the ban.

Cyber Experts Challenge the Ban

A group of 160 leading cybersecurity experts has criticized the ban. They sent an open letter to US Commerce Secretary Howard Lutnick, warning that the ban could harm US and allied cyber defenders. The experts argue that removing access to advanced AI models limits defenders’ ability to address cybersecurity flaws, while adversaries may still access similar or better tools.

The letter states, “To pull the best capabilities away from defenders without a good reason when our adversaries are rapidly advancing is dangerous.” Notable signatories include Alex Stamos, former chief security officer at Facebook; Bruce Schneier, a prominent cybersecurity expert; and Philip Zimmermann, creator of PGP email encryption software.

Concerns About Foreign Access

Reports suggest the ban was motivated by fears that China-linked groups could jailbreak Mythos 5 and Fable 5 and use them for cyberattacks. The experts counter that Chinese AI models are advancing quickly and may already rival US models. “The Chinese open-weight models are only months behind the best American models, and those are the models we know about,” the letter states. The group suggests the Chinese government could have even more advanced, unpublished models.

If these claims are accurate, the ban may not prevent adversaries from accessing powerful AI tools. Instead, it could leave US defenders at a disadvantage.

Capabilities and Limitations of AI Models

Some officials believe these AI models can be easily used for cyberattacks. However, Alex Stamos explains that the process is not simple. He told Axios, “You cannot give Fable the entire Linux kernel and say 'Find all the security bugs.’” Stamos noted that Fable 5’s ability to generate “proof of concept” code for vulnerabilities may have alarmed the White House. Such proofs can help hackers but are also essential for security teams to understand and defend against threats.

Stamos warned, “For us to shut down our best capabilities at the moment we know the Chinese are using and stockpiling these vulnerabilities is dangerous – absolutely foolish. We are in a race right now to fix these bugs as fast as possible.”

Alternative AI Models and Regulation

The ban reportedly followed Amazon researchers’ success in jailbreaking the models using specific prompts. The experts argue that other AI models, such as GPT-5.5, Opus, Sonnet, and Chinese models like Kimi 2.7, can also identify cybersecurity flaws. They state, “AI has been finding bugs and generating working exploits at superhuman levels since last year.”

The group urges that any AI regulation should follow a transparent democratic process, be based on scientific evaluation, and be limited to what is necessary to ensure public safety.

Currently, Mythos 5 and Fable 5 remain unavailable. Anthropic is reportedly in discussions with US officials about possibly reversing the ban.