US blocks Anthropic Claude Fable 5 and Mythos 5 worldwide

The US government has ordered Anthropic to disable access to its Claude Fable 5 and Mythos 5 models for all customers worldwide this week. Anthropic said it is complying with the directive while it works to address the government's concerns, which the order identifies as risks from potential jailbreaks.

What happened

Federal authorities instructed Anthropic to cut global access to the two latest Claude releases, Fable 5 and Mythos 5. The shutdown is immediate and applies to all customers using the models via Anthropic's API, cloud offerings, and any embedded deployments. Anthropic confirmed it has suspended service for those model endpoints and notified affected customers.

The government framed the action around so-called jailbreak vulnerabilities, a class of model behaviors where a system can be coaxed into ignoring safety rules or producing disallowed outputs. The order does not publicly list technical exploits, but it says the risk profile of Fable 5 and Mythos 5 merits removal from public and commercial use until mitigations are in place.

Anthropic said it will cooperate with authorities and pursue measures to restore access where possible. The company also said it is reviewing logs and customer deployments to ensure compliance with the shutdown. Affected customers include cloud platform clients, enterprise API subscribers, and third parties that had integrated the two models into products.

Background and industry context

Claude Fable 5 and Mythos 5 were released as part of Anthropic's latest Claude family upgrades. The models were marketed with expanded capabilities for creative writing, reasoning, and longer-context tasks. Deployers had reported strong performance on several internal benchmarks, which contributed to rapid uptake among startups and enterprises.

Regulators and national security officials have intensified scrutiny of advanced generative models in recent years, focusing on misuse vectors including disallowed content generation, code that could facilitate wrongdoing, and techniques that evade guardrails. The government's action signals a more interventionist stance in cases where it judges a model poses an outsized risk.

Other AI providers have been monitoring the order closely. Some competitors' services remain available, but the enforcement sets a precedent for cross-border reach: a US government instruction has produced a global suspension by a US-headquartered developer.

Immediate impact

Customers relying on Fable 5 and Mythos 5 must migrate to alternative models or earlier Claude versions where available. Companies with live products face urgent remediation tasks: swapping models, revalidating outputs, and communicating changes to users. Smaller developers and startups that built on the two models may face disproportionate disruption, particularly if they lack backup model integrations.

The order will also affect vendors who used the models as embedded components in their software. Those vendors must audit instances, rollback updates that called the disabled endpoints, and coordinate with Anthropic for technical guidance.

Why it matters

A US government order forcing a global model shutdown demonstrates regulators can press vendors to act beyond national markets, introducing operational and legal complexity for AI providers and their customers. Enterprises and developers will likely accelerate contingency planning and diversify model suppliers to avoid single-vendor outages. The decision may also push companies to increase transparency and invest more in defensive testing for jailbreak and misuse scenarios.