US Government Forces Anthropic to Disable Advanced AI Models Over National Security Concerns

The United States Commerce Department forced Anthropic to disable its most advanced artificial intelligence models late Friday evening, marking an unprecedented expansion of national security export controls into the AI sector. The company received a directive at 5:21 p.m. Eastern Time ordering it to immediately suspend access to Fable 5 and Mythos 5 for all foreign nationals, both inside and outside the United States.

The sweeping directive includes foreign nationals residing in America, a category that encompasses Anthropic’s own non-citizen employees. Given the scope of the restrictions, Anthropic argued it had no practical choice but to disable the models entirely for all users globally. The company clarified that access to its less powerful Claude models, including its latest Claude Opus 4.8 variant, remains unaffected by the government order.

Anthropic issued a statement expressing regret for the service disruption. The company stated it believes the situation stems from a misunderstanding and emphasized its commitment to restoring access quickly. The government letter provided no specific details explaining the exact nature of the national security concern that prompted such drastic action.

Company officials revealed that government representatives indicated the decision came after learning of a technique to bypass Fable 5’s safeguards. These protective measures were designed to prevent unauthorized access to the powerful cybersecurity capabilities of Mythos, the underlying AI model on which Fable 5 is built. The government expressed concern that users could exploit this vulnerability to access functions that could be weaponized.

Anthropic Challenges Government Assessment

Anthropic characterized the jailbreak cited by the government as a narrow, specific vulnerability rather than a universal weakness. The company maintained that this particular exploit would unlock Mythos’s cybersecurity capabilities in only one specific instance. It would not defeat all of Fable 5’s safeguards systematically. The company emphasized it had received only verbal evidence of the potential jailbreak without comprehensive technical documentation.

The company asserted that similar jailbreaking methods exist for other publicly available models that face no equivalent export restrictions. Anthropic specifically noted that OpenAI’s GPT-5.5 could be exploited using the same technique. The company stated that jailbreaking methods are widely used in other models. System security professionals routinely employ these methods in legitimate testing scenarios.

“We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to a large number of users,” Anthropic wrote in its blog post.

The company warned that applying this standard uniformly across the AI industry would essentially halt all new model deployments for frontier model providers. Anthropic argued such an approach would create an untenable precedent that would stifle innovation and prevent legitimate technological advancement in the artificial intelligence sector.

Calls for Transparent Regulatory Framework

“As we have stated publicly, we believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts,” Anthropic said. “This action does not adhere to those principles.”

The company’s statement reflects its longstanding position advocating for greater government oversight of AI development. As recently as Wednesday, Anthropic had publicly called for enhanced United States regulation of artificial intelligence systems, including authority to block models with unacceptable risks. However, the company insisted that such oversight must operate through clearly defined legal procedures with proper due process protections.

Industry insiders and policy experts reacted with disbelief to the unprecedented government action. The directive represents the first time the United States has used export controls to restrict access to an AI model itself, rather than the underlying hardware or semiconductor technology that powers such systems.

Escalation of AI Export Control Strategy

The US government has escalated its efforts to control artificial intelligence capabilities that could reach foreign adversaries with this order. For years, American export controls focused primarily on restricting access to advanced chips and manufacturing tools that enable AI development. This new approach directly targets the AI models themselves, marking a significant shift in regulatory strategy.

According to reporting by Axios, government sources indicated the Commerce Department made its decision after a company claimed to have breached Mythos’s security barriers. The government reportedly attempted to temporarily halt Anthropic’s model release through other means but failed, ultimately leading to the issuance of export control notices. These access restrictions will continue for an unspecified period while the government conducts its security review.

Strained Relationship With Trump Administration

The export control directive comes amid already tense relations between Anthropic and the Trump administration. The company’s relationship with the government ruptured earlier this year after it refused to allow the United States military to use its AI models for domestic surveillance and fully autonomous weapons systems. The government responded by placing Anthropic on a supply chain blacklist, with restrictions set to take effect later this year.

The order arrives just as signs suggested the dispute between administration officials and the IPO-bound Anthropic might be easing across parts of the government. The new export control action appears to have reversed any progress toward reconciliation between the company and federal authorities.

Broader Implications for AI Industry

The government directive and Anthropic’s response highlight growing tension between AI developers and regulators over risk assessment methodologies. The incident raises fundamental questions about how governments should exercise authority over AI systems. It challenges assumptions about what constitutes sufficient evidence of security risk. The situation also exposes gaps in current legal frameworks regarding due process for companies facing export control actions.

The precedent set by this case could reshape how frontier AI companies approach model deployment and safety testing. If narrow, specific vulnerabilities become grounds for complete service shutdowns, companies may face impossible standards for demonstrating safety before release. Alternatively, the government action may signal a new era of heightened scrutiny for AI systems with advanced capabilities that could be weaponized or exploited by hostile actors.

Anthropic continues to work with government officials to resolve the situation and restore access to Fable 5 and Mythos 5. The company maintains that transparent dialogue grounded in technical evidence represents the appropriate path forward for AI regulation, rather than abrupt enforcement actions based on incomplete information.