A previously unreleased version of an artificial intelligence (AI) tool, which the company deemed too powerful for public distribution, has now been made available to the public.
Claude Fable 5 represents a variant of Anthropic's Claude Mythos, an AI program that raised significant concerns among technology, financial, and government officials when it was initially released in April for private preview and testing.
Some observers worry the tool is sufficiently advanced to create financial security threats, though others have questioned whether the hype is merely a marketing strategy.
Anthropic announced on Tuesday that Fable will be launched with built-in safeguards and user restrictions, though it acknowledged that "releasing a model this capable comes with risks".
"Fable's capabilities exceed those of any model we've ever made generally available," the company added.
When Mythos was first shared with a limited number of organizations for preview purposes, Anthropic explained it was taking this approach because the tool's high intelligence level could be hazardous due to its capacity to exploit or compromise computer systems.
Canadian Finance Minister François-Philippe Champagne told the BBC in April that the focus on Mythos was justified, partly because "it's the unknown, unknown".
Meanwhile, despite Anthropic's ongoing legal dispute with the US Department of Defense over its refusal to permit government use of its AI tools, US government agencies have also been evaluating Mythos.
Anthropic is anticipated to become a publicly traded company soon, as its private valuation has approached $1tn (£747bn). Consistently demonstrating superior and more extensive AI capabilities strengthens its appeal to investors.
The company also stated on Tuesday that the approximately 150 groups previously granted access to preview Mythos will now receive access to Claude Mythos 5, which lacks restrictions on cybersecurity or biology, depending on each organization's specific applications.
To date, groups and companies utilizing Mythos have reported to Anthropic the discovery of over 10,000 critical security vulnerabilities within their systems.
Anthropic noted that the updated access was restricted to a "small group of cyberdefenders and infrastructure providers" but is expected to expand shortly.
"We intend to expand access to Mythos 5 through a broader trusted access program," the company said.
Both Fable and Mythos, which are fundamentally the same model but with varying safeguards and access levels, can operate "unattended" on human commands given to the tools for extended durations "than any previous Claude models," Anthropic added.
Anthropic co-founder Jack Clark told BBC Newsnight last week that the rapid expansion of AI tool capabilities prompted the company to believe there should be a mechanism for the public to decelerate the technology's progress.
"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."
Finance ministers and bankers raise serious concerns about Mythos AI model
What is Claude Mythos and what risks does it pose?