US Commerce Department Imposes Export Controls on Anthropic AI Models, Citing Jailbreak Risk

The U.S. government has ordered Anthropic to restrict access to its new advanced AI models, Fable 5 and Mythos 5, under export control laws.

This action was triggered by a reported security flaw, known as a 'jailbreak,' which raised national security concerns about potential misuse for cyberattacks.

It marks the first time these regulations have been used to block real-time access to an AI model, setting a major precedent for future AI governance.

MSFT.KSGOOGL.KSAMZN.KSNVDA.KS

The U.S. government has taken the significant step of halting public access to Anthropic's new AI models, Fable 5 and Mythos 5.

This decision didn't come out of nowhere; it's the result of a precise sequence of events. The foundation was laid back in January 2025, when the Commerce Department established new export controls specifically for advanced AI 'model weights'—the core programming of a powerful AI. This rule was combined with a long-standing legal principle called a 'deemed export,' which states that giving a foreign national access to controlled technology, even on U.S. soil, is legally the same as exporting it.

The situation escalated with three key triggers. First, on June 9, 2026, Anthropic launched Fable 5, a very powerful model, making it widely available. This dramatically increased the number of people who could access this frontier technology, raising the stakes for policymakers. Second, just before this, the government was alerted to a security flaw known as a 'jailbreak.' This meant someone had found a way to trick the AI into bypassing its safety rules, creating a potential tool for cyberattacks. Third, this specific risk aligned perfectly with a new Executive Order signed on June 2, which prioritized national security and cyber threats related to AI.

These factors led to the climax on June 12. The Commerce Department used the 2025 rule to officially classify Fable 5 and Mythos 5 as controlled technology. Citing the 'deemed export' doctrine, they argued that allowing any non-U.S. person to use the models was an illegal export. To comply with this directive, Anthropic had no choice but to shut down access for everyone, everywhere. This event marks a major shift from voluntary AI safety guidelines to mandatory, legally enforced action, sending a clear message to the entire industry: credible security threats can and will trigger a swift government shutdown.

Glossary
Export Controls: Government rules that restrict the transfer of sensitive technology to other countries or foreign nationals for national security reasons.
Deemed Export: A legal rule that treats the release of controlled technology to a foreign person within the U.S. as an export to that person's home country.
Jailbreak: The act of using carefully crafted prompts to bypass an AI model's safety features, making it perform restricted or harmful tasks.