Anthropic Shuts Down 2 Frontier AI Models Over Jailbreak Fears After US Order

For AI researchers and engineers, this incident underscores the fragility of AI alignment measures and the real-world consequences of jailbreak discovery. It raises critical questions about how safety evaluations should be conducted and shared between public sector, private labs, and government agencies, potentially reshaping AI development practices.

On June 12, 2026, the US Commerce Department issued an export control directive ordering Anthropic to block foreign national access to its most advanced AI models, Fable 5 and Mythos 5. The order forced the company to immediately disable both models for all customers — a dramatic escalation in Washington's efforts to control frontier AI technologies. The move was prompted by a reported 'jailbreak' — a method to bypass the model's safeguards against exploiting software vulnerabilities — that another company allegedly demonstrated to the government. Anthropic, a darling of AI safety circles, pushed back publicly, calling the evidence 'narrow' and stating the capability is already widely available in other models. This episode marks the first time the US has restricted a commercial AI model through export controls, shifting from the previous focus on semiconductor chips and manufacturing tools.

“

On June 12, 2026, the US Commerce Department issued an export control directive ordering Anthropic to block foreign national access to its most advanced AI models, Fable 5 and Mythos 5.

The timing is especially sensitive for Anthropic, which confidentially filed for an IPO in May 2026 and was already in a dispute with the Trump administration after refusing to allow its AI to be used for domestic surveillance and fully autonomous weapons. That earlier standoff landed Anthropic on a supply chain blacklist. Now, the company must disable products that it had just launched — Fable 5 only debuted earlier that week — potentially impacting hundreds of millions of users and its revenue trajectory. The Pentagon's CIO, Kirsten Davies, reinforced the government's stance, stating on social media that national security outweighs 'revenue cycles, clickbait, and pre-IPO valuation.'

The international fallout is immediate. EU institutions and non-US allies that rely on Anthropic's models are scrambling to assess the impact. The UK's AI minister, Kanishka Narayan, emphasized the need for technological sovereignty, underscoring a growing unease among US allies about dependency on American AI. This could accelerate efforts in Europe and elsewhere to develop independent AI capabilities.

From a legal standpoint, the order raises significant due process concerns, as Anthropic was given only verbal evidence and no formal administrative hearing before being forced to disrupt its commercial operations. The company's disagreement with the government's rationale may pave the way for litigation over the scope of export control authority under the Export Control Reform Act or similar statutes. It also tests the balance between national security and innovation, with critics arguing that overly broad restrictions could hinder US competitiveness if models are crippled while foreign alternatives thrive.