Anthropic has suspended access to its newly launched Claude Fable 5 and Mythos 5 AI models for users outside the United States after receiving an export control directive from the US government.
The move comes just days after the company launched Fable 5 and Mythos 5, which are among Anthropic’s most advanced AI models to date. According to the company, the order was issued under US national security authorities and requires access to the two models to be restricted for all foreign nationals, regardless of whether they are located inside or outside the United States.
As a result, users in countries such as India, the United Kingdom, Canada, Australia, and most other regions can no longer access the new models. However, access for US citizens and authorized users within the United States remains unaffected.
Anthropic revealed that it received the directive from the US government at 5:21 p.m. ET. The company said the order did not include details about the specific national security concern that led to the restriction.
The company believes the government’s concerns may be related to reports of a possible jailbreak technique capable of bypassing some of Fable 5’s safeguards.
A jailbreak refers to a method used to trick an AI model into ignoring its safety restrictions and generating responses that would normally be blocked.
However, Anthropic stated that despite extensive testing, no researcher or security team has successfully discovered a universal jailbreak that consistently bypasses Fable 5’s protections.
Anthropic launched Claude Mythos Preview earlier this year for a limited number of organizations. The model was designed for advanced coding, cybersecurity, and scientific applications.
Earlier this week, the company introduced Fable 5 and Mythos 5 to a broader audience.
Fable 5 was designed as a general-purpose AI model with strict safeguards, while Mythos 5 targeted specialized use cases in cybersecurity and biology, where users often require fewer restrictions to perform advanced research and analysis.
Anthropic says both models underwent extensive safety testing before launch. The company worked with government agencies, independent organizations, and internal teams to evaluate potential risks and vulnerabilities.
The company has openly questioned the government’s decision.
Anthropic argued that a potential jailbreak should not be enough reason to restrict access to a commercial AI model. The company warned that if similar standards are applied across the AI industry, it could significantly slow the deployment of future frontier AI systems.
The company is currently working with authorities and hopes to restore access for foreign users as soon as possible.
The decision could have implications for India, which is one of the world’s largest markets for AI tools and services.
Indian developers, startups, enterprises, researchers, and cybersecurity professionals who were hoping to use Fable 5 or Mythos 5 may now be unable to access the models despite being willing to pay for them. This shows a growing challenge for countries that rely heavily on AI technologies developed in the United States.
While Anthropic’s older models remain available, the restriction means Indian users could miss out on the latest capabilities offered by the company’s newest systems.
Historically, export controls have been used for technologies such as advanced semiconductors, military systems, and high-performance computing hardware. If advanced AI models are now being treated in a similar way, it could be the beginning of a new era where access to frontier AI depends not only on technical capability but also on government policy.
Now India should also invest more heavily in domestic AI development. Indian startups and government-backed initiatives are already working on local AI models, and incidents like this may strengthen the argument for building independent AI capabilities.
Related Posts
