Anthropic halts Fable 5 and Mythos 5 after U.S. export-control order
Anthropic has suspended access to its newly launched Fable 5 and limited-release Mythos 5 models after receiving a U.S. government export-control directive citing national-security concerns.
What happened
- Anthropic says it received the directive at 5:21 p.m. ET ordering the company to block access to Fable 5 and Mythos 5 for all foreign nationals—whether inside or outside the U.S.—including foreign-national employees.
- To comply immediately, Anthropic disabled both models for all users. The company said other models, including Opus 4.8, remain available and are not subject to the restriction.
- The action follows the recent rollout: Fable 5 was just made generally available as a Mythos-class model, while Mythos 5 had been released to a limited set of approved cybersecurity and infrastructure users.
Why the U.S. intervened
- Authorities have pointed to a possible “jailbreak” technique that could bypass some safeguards. Anthropic says officials have provided only verbal, limited evidence describing a narrow, non‑universal jailbreak.
- According to Anthropic, the reported method involves asking the model to analyze a specific codebase and identify or fix software vulnerabilities—a task that differs from a universal jailbreak that would broadly strip safety protections across many tasks.
- Anthropic argued that treating a narrow, situational vulnerability as grounds for recalling a commercial model used by millions would set an industry‑wide precedent that could stifle new model deployments.
Anthropic’s response and next steps
- The company says it is cooperating with authorities, believes the directive may stem from a misunderstanding, and is working to restore access as quickly as possible.
- Meanwhile, Anthropic continues to expand its compute capacity for future systems. Private credit firms Blackstone and Apollo Global Management are reported to be syndicating roughly $36 billion to back Anthropic’s next round of infrastructure spending. Reuters says the financing will help Anthropic acquire custom tensor processing units from Google—supported by Broadcom technology—that the company plans to lease for its AI operations.
Policy push and broader context
- Anthropic has been actively pushing governments to establish rules for frontier AI, proposing policies on dangerous deployments, independent evaluation, cybersecurity safeguards, and economic measures for workers affected by AI. The new government intervention marks one of the most significant regulatory moves involving a recently released frontier model.
Why crypto and Web3 watchers should care
- The order has implications for developers and auditors who use cutting‑edge models for code review, smart-contract security, and vulnerability analysis. A temporary or prolonged cutoff of high-capability models like Fable 5 could slow security workflows in crypto and other software-focused sectors.
- At the same time, the large infrastructure financing underlines investor confidence in AI compute capacity—an important factor for projects that combine AI and blockchain services or rely on on‑demand model access.
Anthropic’s pause puts it squarely at the intersection of rapid AI advancement and emerging national-security oversight—an episode likely to shape how future frontier models are deployed and regulated.
Read more AI-generated news on: undefined/news
