US halts access to Anthropic's Fable 5 and Mythos 5 models over security concerns
The US government has abruptly suspended global access to Anthropic’s Fable 5 and Mythos 5 AI models, citing national security concerns—but the reasoning remains unclear. In a letter sent to the company, the Commerce Department ordered the immediate halt for all foreign nationals, including Anthropic employees, without providing specific technical justification. To comply, Anthropic disabled both models entirely rather than risk violating the directive.
A disputed rationale
Anthropic disputes the government’s assessment, arguing the decision lacks transparency and is not grounded in clear evidence. The company claims it was told a narrow jailbreak technique—one that could prompt a model to identify software vulnerabilities in a codebase—triggered the restriction. However, Anthropic disputes the severity of the risk, stating that similar techniques are already used widely across the industry without triggering such action. It also notes that no AI system can be made completely jailbreak-proof and that its own layered security measures, including 30-day data retention for threat detection, were designed to mitigate such risks.
Industry implications
The move arrives as Anthropic disclosed a $47 billion revenue run rate and a $965 billion valuation in a recent IPO filing, making the timing particularly sensitive. While other Anthropic models remain unaffected, the blanket suspension raises questions about the government’s approach to AI oversight. Anthropic argues that applying such a standard across the industry could halt new model deployments entirely, calling for a transparent, fact-based process for evaluating risks. The dispute highlights the growing tension between innovation and regulation in the AI sector.
Source: Security Affairs. AI-assisted editorial synthesis — TechnoExpress.