Anthropic has been compelled to suspend access to its flagship AI models, Fable 5 and Mythos 5, following a US government export-control directive issued on national-security grounds.

The company said it received the order at 5:21 p.m. ET, requiring the immediate suspension of access to the two models for all foreign nationals, regardless of whether they are located inside or outside the United States. The directive also applies to foreign-national employees working at Anthropic. To ensure compliance, Anthropic has disabled access to both models for all customers globally. Access to the company's other AI models remains unaffected.

According to Anthropic, the government did not provide detailed information regarding the specific national security concerns behind the action. However, the company believes the directive is linked to allegations that a method exists to bypass, or "jailbreak," the safety mechanisms embedded within Fable 5.

Anthropic said it reviewed demonstrations of the reported technique and concluded that it exposed only a small number of previously known and relatively minor vulnerabilities. The company further stated that similar vulnerabilities can be identified by other publicly available AI models without requiring any safety bypass.

The move marks one of the most significant government interventions in the deployment of advanced AI models and raises questions about how regulators will balance national security concerns with the continued advancement of frontier AI technologies.

Anthropic Defends Safety Measures

The company strongly defended the safety architecture of Fable 5, highlighting extensive pre-release testing conducted in collaboration with US and UK government agencies, independent third-party organisations and internal red teams.

According to Anthropic, thousands of hours of evaluations were performed before launch, with results indicating that Fable 5's cybersecurity safeguards were substantially stronger than those of previously deployed AI systems.

Anthropic noted that no testers had identified a "universal jailbreak" capable of broadly bypassing the model's protections across a wide range of cyber-related tasks. Instead, the company acknowledged that narrow, non-universal jailbreaks may occasionally expose limited information, a challenge it argues is common across the AI industry.

The company also emphasised that it had implemented a defence-in-depth approach, combining layered safeguards with monitoring systems and a 30-day data retention policy designed to detect and mitigate emerging threats.

Dispute Over Regulatory Threshold

Anthropic disagreed with the government's decision to remove the models from the market, arguing that the reported vulnerability does not justify a global shutdown of a commercial AI system.

The company stated that the disclosed technique essentially involved asking the model to review software code and identify potential flaws, a capability it claims is widely available across competing frontier models and routinely used by cybersecurity professionals.

Anthropic warned that applying such a standard across the industry could significantly hinder the deployment of future frontier AI models.

"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," the company said. "This action does not adhere to those principles."

Broader Implications for AI Governance

The suspension comes amid increasing global scrutiny of advanced AI systems and growing debates over the role governments should play in regulating powerful foundation models.

The case highlights the emerging tension between national security oversight and commercial AI innovation, particularly as governments seek to address risks posed by increasingly capable AI systems while ensuring that regulatory actions remain transparent and technically justified.

Anthropic said it is complying with the directive while working with authorities to resolve what it described as a misunderstanding and restore access to Fable 5 and Mythos 5 as quickly as possible.