Anthropic has said it will disable access to its advanced AI models Fable 5 and Mythos 5 for all users after receiving a US government order restricting access for foreign nationals. The directive cited national security concerns, but the company said it was not given detailed evidence explaining the specific threat.
The decision affects two of Anthropic’s most powerful AI systems and comes shortly after the launch of Claude Fable 5, which the company described as part of a higher-capability Mythos-class model tier. While the US order was aimed at foreign nationals, Anthropic said it must switch off access for all customers to ensure compliance.
The move marks a major escalation in US AI export control policy. For years, restrictions focused mainly on advanced chips, semiconductor tools and computing infrastructure used to train AI models. This directive goes further by targeting access to the AI models themselves, raising new questions for the global AI industry.
US Raises AI Jailbreak Concern
According to Anthropic, the government believes there may be a way to bypass safeguards in Fable 5. Such bypass methods, often called jailbreaks, can allow users to push AI systems outside their safety limits. In this case, the concern reportedly involves the possibility of using the model to identify software vulnerabilities.
Anthropic said the government provided only verbal evidence of a narrow and non-universal jailbreak risk. The company disagreed with the order, arguing that a limited potential issue should not justify recalling a commercial model already deployed to a large user base.
The company also said it had worked with the US government and other groups on safety measures before releasing Fable 5. It added that rival AI providers have models with similar abilities to find minor bugs in code, suggesting that the concern may not be unique to Anthropic.
AI Regulation Debate Intensifies
The dispute highlights growing tension between frontier AI companies and regulators over how to manage fast-evolving security risks. Governments are increasingly worried that advanced AI models could be misused for cyberattacks, especially against banks, defence networks, critical infrastructure and older technology systems that may contain hidden weaknesses.
Anthropic said access to its other AI models will not be affected. However, the shutdown of Fable 5 and Mythos 5 could still have major consequences for developers, businesses and enterprise users who had begun testing the new models.
Amazon Web Services said Anthropic had asked it to revoke access to the affected models for all users in all regions. A US official also confirmed that the Commerce Department issued an export control directive restricting foreign national access to the models.
The issue arrives at a sensitive time for Anthropic, which has reportedly filed confidentially for a US IPO. The company’s relationship with parts of the US government had already faced strain after disagreements over military use, domestic surveillance and fully autonomous weapons systems.
Anthropic said it believes there has been a misunderstanding and is working to restore access as soon as possible. But the sudden shutdown has sent a clear signal: advanced AI models are now becoming part of the same national security debate that already surrounds chips, cloud infrastructure and cyber defence.
For the wider AI industry, the case could become a turning point. If similar standards are applied across all frontier AI companies, future model launches may face tougher government review, stricter user verification and possible restrictions based on nationality or export-control rules.
