Anthropic disabled access to its Claude Fable 5 and Mythos 5 artificial-intelligence models after receiving a US government directive tied to export controls and national-security concerns. The company said the order required it to prevent foreign nationals from using Fable 5, prompting a broader shutdown that affected all customers while it sought to comply.
The intervention came only days after Fable 5 became publicly available. Anthropic had promoted the model as its most capable general release while also emphasizing safeguards intended to constrain its potential use in cyberattacks. Before the public launch, a small group of organizations received access to test the system and examine possible vulnerabilities.
According to Anthropic's account, US authorities had not identified a specific security flaw. The company said it understood that officials were concerned about a technique for bypassing Fable 5's restrictions. Anthropic reviewed a demonstration and said it revealed a small number of minor vulnerabilities that were already known. It also maintained that publicly available rival models could identify the same weaknesses without such a bypass.
The dispute highlights the tension between rapidly improving AI capabilities and efforts to control access to tools that may assist cybersecurity work. Gina Neff, a professor of responsible AI at Queen Mary University London, warned that restricting a model could also impede safety testing and international cooperation. She cited testing by the UK government's AI Security Institute in which the model was able to exploit defenses and systems 73% of the time, describing that performance as a significant shift in cyber capability.
The European Union, which had gained access to Mythos earlier in June following talks, said it was assessing Anthropic's announcement. European Commission spokesman Thomas Regnier said the episode reinforced the case for European technological sovereignty. The bloc had recently announced measures intended to reduce its reliance on American and Asian suppliers for important technologies, including AI.
The shutdown unfolded alongside a separate confrontation between Anthropic and the Trump administration. President Donald Trump had publicly criticized the company, and Defense Secretary Pete Hegseth designated it a supply-chain risk, a label indicating that a service is considered insufficiently secure for government use. Anthropic sued the Pentagon over that designation. A judge subsequently ruled that the directive could not be enforced while the case continued, leaving government agencies and military contractors able to use Anthropic tools during the litigation.
The Commerce Department had not supplied the BBC with a response at the time of its report. Anthropic's decision therefore left open how long the models would remain unavailable and whether narrower access controls could eventually replace the company-wide suspension.
