Anthropic’s AI Model Mythos 5 Returns with US Government Restrictions After Two-Week Ban

After two weeks of tense negotiations with the Trump administration, Anthropic’s powerful AI model "Mythos 5," which had been blocked shortly after its public release, has been authorized for limited redistribution. The model will initially be available only to a select group of about 100 American organizations and companies, primarily those involved in cybersecurity and infrastructure. However, the public version "Fable 5" remains under an uncertain status with no clear timeline for broader release.

Anthropic reported that the government was presented with a cybersecurity vulnerability related to Fable 5, though the company described these issues as relatively minor. The firm also warned that other advanced AI models might raise similar concerns. In a formal letter sent two days ago, Commerce Secretary Howard Lutnick informed Anthropic co-founder Tom Brown, who led the negotiations, that the decision followed the company’s implementation of fixes meeting licensing requirements. Lutnick noted that Anthropic has recently collaborated with the US government to address risks associated with both models.

Anthropic spokesperson Danielle Giliri stated the company received official notice that Mythos 5 could be shared with a limited group of defensive cybersecurity firms and infrastructure providers. She expressed optimism about expanding access to Mythos 5 and eventually restoring Fable 5 for general use. Nevertheless, the government’s directive barring foreign nationals, including Anthropic employees, from accessing the models remains in place. This approval is thus an exceptional allowance for a restricted group, with no indication when or if the restrictions will be lifted.

Simultaneously, OpenAI announced that after a two-week distribution ban, it also received limited government approval to release its new GPT 5.6 model to a pre-approved set of organizations, including government bodies, infrastructure companies, and trusted enterprise clients. OpenAI criticized the new government approval process, warning it delays advanced AI tool deployment to developers, organizations, and cybersecurity firms. The company emphasized ongoing cooperation with the government, hoping for eased restrictions and broader future releases. OpenAI described the approval process as a temporary step toward establishing a permanent regulatory framework for future AI models.