In a move that has stirred debate across technology, finance, and government circles, Anthropic – the AI start‑up behind the contentious Claude Mythos platform – has released its newest iteration, Claude Mythos 5, to the public. The company, which had earlier warned that the model was “too powerful for public deployment,” announced on Tuesday that safeguards and user limits would be put in place, stating that “releasing a model this capable comes with risks.”
“Fable’s capabilities exceed those of any model we've ever made generally available,” the organisation said, referring to the underlying technology that powers both Mythos 5 and its private testing counterpart, Claude Fable 5.
“We intend to expand access to Mythos 5 through a broader trusted access program,” Anthropic added, hinting that the model will eventually be offered beyond the handful of cyberdefenders and infrastructure providers currently allowed to use it.
The story first emerged during a private preview of Mythos in April, when Tatiana P. Smith, a senior analyst at TechScan, reported that the model was “intelligent enough to exploit and potentially hack systems.” Canadian Finance Minister François‑Philippe Champagne echoed this worry by describing the AI as “the unknown, unknown” and a risk to financial security.
Despite an ongoing lawsuit against the U.S. Department of Defense over its refusal to adopt Anthropic’s tools, U.S. agencies have quietly begun testing Mythos. In the meantime, the company now reports that re‑examining the systems of its early adopters uncovered more than 10,000 critical security flaws – a figure that underscores the necessity of stringent controls.
Jack Clark, Anthropic co‑founder, recently told BBC Newsnight that AI’s growing pace has left the industry with a gas pedal but no brake, and that “the option to take your foot off the gas and put your foot on the brake” is essential.
With its public valuation approaching a trillion dollars and plans to go public soon, Anthropic’s bold leap represents a gamble on the future of AI, weighing potential economic gains against the real threat posed by a model that could push the boundaries of current defensive mechanisms.
