The U.S. government has directed artificial intelligence company Anthropic to disable its most advanced AI models following concerns about potential security vulnerabilities. The directive represents a significant intervention in commercial AI deployment and highlights growing tension between innovation and national security.
Anthropic received an export control order from the U.S. government requiring the company to suspend access to its Fable 5 and Mythos 5 models for all foreign nationals. The government cited national security concerns, though Anthropic stated it was not provided specific details about the threat. According to the company, the government believes safeguards in these models can be bypassed and the tools could potentially be used to identify software vulnerabilities.
Rather than restrict access only to foreign users, Anthropic announced it would "abruptly disable" the models for all users globally. The decision affects advanced AI systems that had recently been released to the public.
The shutdown stems partly from a jailbreak discovered in Fable 5, which the government views as a potential security risk. Anthropic has expressed clear disagreement with the government's response. In a statement, the company wrote: "We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people."
The disagreement underscores a fundamental difference in how Anthropic and government officials approach AI safety. Anthropic emphasizes its safety-focused development practices, including testing for vulnerabilities before releasing models to the public. The company argues that discovering a narrow jailbreak in an already widely deployed system does not justify a complete shutdown.
The Commerce Department maintains authority over technology exports and emerging technologies affecting national security. This enforcement action signals that federal agencies are increasing their scrutiny of AI systems and their potential security implications. As advanced AI technology becomes more powerful and widely deployed, regulators are grappling with how to balance innovation against security risks.
The case marks a notable moment in government regulation of artificial intelligence. Companies developing advanced AI systems now face mounting pressure to address security vulnerabilities while navigating evolving regulatory requirements. Federal agencies continue developing frameworks for AI oversight, and this action may establish precedent for how the government responds to discovered vulnerabilities in widely deployed systems.
The decision also raises questions about the process for such interventions. Anthropic's statement that it received no specific details about the national security concern suggests potential friction between the company and regulators over transparency in enforcement actions. As AI technology becomes increasingly central to business and infrastructure, how agencies communicate their security concerns may prove as important as the concerns themselves.