Anthropic releases public version of “Claude Mythos” model, after privately previewing it for months

Anthropic has released Claude Fable 5, a publicly available version of its Claude Mythos AI system, reversing earlier efforts to restrict access to the model. The company had described Mythos as too capable to be broadly released, after a limited preview of the technology drew attention from leaders in finance, technology, and government.

According to reporting, Anthropic rolled out Mythos privately in April for previewing and testing. During that earlier phase, the company said it was limiting availability because of the system’s ability to exploit or hack computer systems, and because “releasing a model this capable comes with risks.” In the release described Tuesday, Anthropic said it is now making the Fable 5 version available, with safeguards and user limitations in place.

The earlier preview was shared with roughly 150 organizations, which Anthropic said will now be given access to Claude Mythos 5 (described in reporting as a follow-on version associated with the same technology effort). The company’s statement, as reported, also indicated that Fable’s capabilities exceed those of any model it had previously made generally available, while maintaining that releasing such systems requires additional controls.

The Mythos preview drew direct attention from officials who framed the issue as one of uncertainty and potential misuse. Canadian Finance Minister François-Philippe Champagne told the BBC in April that the concern was partly driven by “the unknown, unknown,” describing uncertainty around what capabilities could be used for and how quickly they could be applied outside controlled testing environments.

While Anthropic faced public scrutiny over its approach, government bodies were also reported to have tested Mythos. The BBC report states that U.S. government agencies had been testing Mythos even as the company pursued legal action against the U.S. Department of Defense, over the department’s refusal to accept government use of Anthropic’s AI tools.

Alongside the release, the company’s plans for its corporate future were also part of the broader context. The reporting indicates Anthropic is expected to become a public company soon, with a private valuation nearing $1 trillion, and that demonstrating broader and improved AI capabilities is relevant to investor expectations.

In the immediate term, the public release shifts Mythos from a limited preview to a wider user base, increasing the number of people and organizations that could potentially interact with the model. Anthropic’s stated use of safeguards and limitations is expected to be the key mechanism for reducing misuse risk, and the company will likely face further scrutiny from regulators and governments as more users gain access.

For now, the practical next steps for users and institutions will hinge on the conditions attached to Claude Fable 5 access, as well as on how effectively the safeguards prevent cybersecurity exploitation and other prohibited uses that earlier officials and the company itself highlighted as concerns.