Anthropic released Claude Fable 5 on Tuesday, putting a public-facing version of its powerful Claude Mythos AI model into general circulation months after the company warned the same underlying technology could pose serious cybersecurity risks if released without controls.
The company acknowledged the tension in a statement. “Releasing a model this capable comes with risks,” Anthropic said. “Fable’s capabilities exceed those of any model we’ve ever made generally available.” The company said it had equipped Fable with safeguards and user limitations before the wider release.
Claude Mythos, the underlying model, was first made available in April to a small group of organizations for preview and testing. At the time, Anthropic said it was restricting access because the tool was intelligent enough to be capable of autonomously exploiting or hacking computer systems. The caution attracted attention from technology, finance, and government leaders, some of whom warned the tool could pose financial security risks.
Canadian Finance Minister François-Philippe Champagne told the BBC in April that he considered the attention on Mythos warranted, in part because of what he described as “the unknown, unknown.”
The roughly 150 groups that had been given access to preview Mythos will now receive access to a more capable version, Claude Mythos 5, which does not impose limitations on cybersecurity or biology-related applications depending on each organization’s specific use case, Anthropic said.
Anthropic said those organizations have collectively reported finding more than 10,000 critical security flaws in their systems using Mythos. The company said it is keeping the updated access limited to “a small group of cyberdefenders and infrastructure providers” for now but intends to expand access through a broader trusted access program.
Both Fable and Mythos — described by the company as essentially the same model with different safeguards and levels of access — can operate “unattended” on human commands for longer periods than any previous Claude models, Anthropic said.
The release comes as Anthropic’s private valuation approaches $1 trillion, and the company is expected to file for an initial public offering. The company is also locked in a continuing legal dispute with the U.S. Department of Defense after the Pentagon designated Anthropic a supply chain risk and barred it from defense contracts earlier this year, following collapsed talks over military use of the company’s AI tools. Anthropic sued the Trump administration in March to undo that label; an appeals court rebuffed the company in April in a related fight.
Despite the lawsuit, U.S. government agencies have been testing Mythos as well, according to the BBC.
Anthropic co-founder Jack Clark told BBC Newsnight last week that AI capabilities were expanding so rapidly the industry should provide the public a mechanism to slow development. “You want the option to be able to take your foot off the gas and put your foot on the brake,” Clark said. “Right now, it’s like the AI industry has a gas pedal, but it doesn’t have a brake pedal.”