Anthropic has disabled access to its most powerful artificial intelligence models, Claude Fable 5 and Claude Mythos, in response to a directive from the Trump administration. The government order restricts foreign users from accessing these advanced AI systems on national security grounds.
The move comes after unauthorized users gained access to Claude Mythos through a third-party vendor environment. Bloomberg reported that a small group of people in a private online forum obtained access to the model, which had not been publicly released due to concerns about its potential to enable cyberattacks. One member of the group worked as a contractor for Anthropic, and they used methods employed by cybersecurity researchers to gain entry.
According to Bloomberg's reporting, the unauthorized users were primarily interested in exploring the technology rather than causing harm. They did not attempt to run cybersecurity attacks on the model. The publication verified their claims through screenshots and a live demonstration.
The incident raises serious concerns among government officials and security experts. Kanishka Narayan, the UK's AI minister, warned that UK businesses "should be worried" about Mythos's capability to identify flaws in IT systems that hackers could exploit. The UK's AI Security Institute, considered the world's leading safety authority on AI technology, previously warned that Mythos represented a significant step forward in terms of cybersecurity threats.
The institute found that Mythos could execute complex attacks requiring multiple steps and discover IT system vulnerabilities without human assistance. Tasks that would normally require days of work by human professionals could be completed by the model. Notably, Mythos became the first AI model to successfully complete a 32-step cyberattack simulation created by the institute, solving it in three of ten attempts.
Anthropic expressed disagreement with the decision to halt the models. The company wrote in a statement that it "disagrees that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people."
The Trump administration's directive reflects broader efforts to control international access to cutting-edge American AI technology amid intensifying global competition in artificial intelligence development. The order prioritizes American technological dominance while addressing potential security vulnerabilities in advanced AI systems.
Anthropic's decision to comply with the government directive demonstrates how regulatory pressure and security concerns are shaping the deployment of powerful AI models. The situation highlights ongoing tensions between rapid AI development and safety precautions, as both industry leaders and government officials grapple with how to manage potentially dangerous capabilities in increasingly sophisticated systems.