In a move that has reignited global debate over the risks and rewards of cutting-edge artificial intelligence development, AI startup Anthropic has announced the public release of Claude Fable 5, a model the firm itself previously deemed too powerful to share with the general population.
The release comes nearly three months after the company first rolled out a private preview of Claude Mythos, the base architecture for both new models, to a select group of testing organizations in April. That early limited release sparked immediate alarm across technology, financial, and government circles, with many stakeholders flagging the model’s unprecedented capabilities as a major potential threat to digital and economic security. Critics have also pushed back, however, arguing that much of the hype surrounding the model’s power is little more than deliberate marketing positioning to boost the company’s profile ahead of its expected public listing.
When Anthropic first shared Mythos with its small initial test group, company leaders openly warned that the model’s advanced intelligence gave it the ability to exploit vulnerabilities and hack computer systems, making it inherently dangerous for broad distribution. Canadian Finance Minister François-Philippe Champagne echoed that uncertainty in an April interview with the BBC, noting that heightened scrutiny of the model was justified by the sheer scale of uncharted risk it represented, calling it an “unknown, unknown.”
Notably, even amid an ongoing legal dispute between Anthropic and the U.S. Department of Defense over the company’s refusal to allow its AI tools to be used for government purposes, multiple U.S. federal agencies have already joined the early testing program for Mythos.
The company’s upcoming initial public offering is a major context for this release: Anthropic’s current private market valuation has already climbed to nearly $1 trillion (£747 billion), and demonstrating consistent, expanding AI capabilities is a key step to reinforce its appeal to prospective public investors.
Alongside the public launch of Fable 5, Anthropic announced Tuesday that the roughly 150 organizations that participated in the original Mythos preview will now gain access to Claude Mythos 5. Unlike Fable, Mythos 5 does not include built-in restrictions related to cybersecurity and biological research use cases, with access tailored to an organization’s specific authorized activities. To date, organizations testing early versions of the model have reported that it helped them identify more than 10,000 critical security flaws in their internal systems, a tangible benefit that company leaders highlight to justify expanding access.
Right now, expanded access to Mythos 5 is restricted to a “small group of cyberdefenders and infrastructure providers,” but Anthropic confirmed it plans to roll out access more widely in the near future through a formal trusted access program for vetted organizations.
Both Fable 5 and Mythos 5 are built on the same core model architecture, differing only in the safeguards and access restrictions applied to each. Anthropic confirmed that both models are capable of operating autonomously “unattended” to complete complex user commands over much longer time frames than any previous iteration of the company’s Claude models.
Even as the company moves forward with releasing these more capable models, Anthropic co-founder Jack Clark warned last week in an interview with BBC Newsnight that AI capabilities are advancing so quickly that the industry needs to build guardrails to allow for slowing development if needed. Clark argued that the current AI ecosystem is structured to push for constant acceleration without any mechanism to pause or slow progress, saying, “You want the option to be able to take your foot off the gas and put your foot on the brake. Right now, it’s like the AI industry has a gas pedal, but it doesn’t have a brake pedal.”
Anthropic emphasized in its announcement Tuesday that Fable 5 is being launched with a full suite of built-in safeguards and user limitations in place, but the company did not downplay the inherent risks of releasing a model of this capability. The firm acknowledged openly that “releasing a model this capable comes with risks,” adding that Fable’s capabilities outpace any model the company has ever made available to the general public.