US Government Orders Anthropic to Halt Access to Claude Fable 5 and Mythos 5 Models
The U.S. government has ordered Anthropic to immediately suspend access to its most advanced artificial intelligence models, Claude Fable 5 and Claude Mythos 5, citing national security concerns. According to a report from TechCrunch, the company complied with the order on Friday evening, though it publicly contested the necessity of such a restrictive measure.
Why did the U.S. government restrict these AI models?
Federal authorities identified potential security risks related to “jailbreaking,” a technique used to bypass developer-imposed safety protocols. Anthropic confirmed through its company blog that the government expressed concerns regarding the models’ ability to assist in high-risk areas like cybersecurity and biology. While Fable 5 was designed with “guardrails” to prevent the generation of harmful content, officials argued that the potential for these safeguards to be circumvented justified a total suspension of service.
Claude Mythos 5 was originally restricted to a select group of only 50 organizations, including Apple, Google, and Microsoft, specifically for use in the “Project Glasswing” cybersecurity program.
How does the industry view the risk of “jailbreaking”?
Anthropic argues that the government’s standard for restriction is disproportionate to the actual risk. The company maintains that the ability to analyze code for vulnerabilities—a core feature of Mythos 5—is already present in other publicly available models, such as OpenAI’s GPT-5.5. Anthropic points out that cybersecurity experts have utilized these capabilities for years to defend digital infrastructure. According to the company, independent “classifier” systems are already in place to filter outputs, providing a secondary layer of protection even if a user attempts to bypass initial guardrails.
What are the implications for AI development?
The forced withdrawal of a commercial model used by millions sets a potentially disruptive precedent for the industry. Anthropic stated that if the government’s current criteria for “limited jailbreak potential” become the industry standard, it could stifle the release of all future high-capability AI models. This tension highlights a growing divide in the tech sector: how to balance rapid innovation with the government’s mandate to regulate powerful, potentially dual-use technologies.
Comparison: Security Features and Access
| Model | Primary Purpose | Access Level |
|---|---|---|
| Mythos 5 | Vulnerability detection | Restricted (50 orgs) |
| Fable 5 | General use with guardrails | Suspended (Global) |
Frequently Asked Questions
- Are all Anthropic models suspended? No. Only the Fable 5 and Mythos 5 models were affected by the government’s recent directive. Other models remain accessible.
- What is a jailbreak in AI? It refers to techniques used to bypass the safety rules or behavioral constraints programmed into an AI model by its developers.
- Why does the government care about code analysis? Because models capable of finding vulnerabilities in software can be used by malicious actors to launch cyberattacks, despite their intended use for defensive security.
What do you think about the balance between AI safety and innovation? Share your thoughts in the comments below or join the discussion on our community forum.