A highly advanced artificial intelligence system once considered too powerful for public release has now been made available, marking a significant development in the rapid evolution of AI capabilities and raising renewed debate over safety, security, and oversight.
The model, called Claude Fable 5, is a public-facing version of Anthropic’s earlier system Claude Mythos, which was previously shared only with a limited group of organisations for testing. At the time of its private release in April, the model drew attention from technology experts and policymakers who warned about its potential to be misused in areas such as cybersecurity and financial systems.
Anthropic said the original Mythos model demonstrated capabilities that exceeded any of its previous publicly available systems. The company said concerns were raised internally about whether such a tool could be used to exploit computer networks or assist in hacking activities, prompting restrictions on access during early testing.
In a statement on Tuesday, the company confirmed that Fable 5 would now be released to the public with safety guardrails and usage limits in place. However, it acknowledged that deploying such a capable system carried inherent risks. The company also said the model represents its most advanced AI system to date.
Claude Mythos was initially made available to around 150 organisations, including cybersecurity specialists and infrastructure providers. Those groups reportedly used the system to identify more than 10,000 critical vulnerabilities across various digital systems, highlighting its potential usefulness in defensive security work.
Anthropic said access to Fable 5 will initially remain limited, with expanded availability planned through a controlled programme. Certain technical restrictions, including safeguards around cybersecurity and biological applications, will vary depending on the organisation using the system.
The company also noted that both Fable and Mythos models are capable of operating autonomously for extended periods, executing complex tasks based on user instructions without continuous human supervision. Anthropic described this extended operational ability as more advanced than any previous version of its Claude AI family.
Concerns over the rapid development of such systems have also been echoed outside the company. Canadian Finance Minister François-Philippe Champagne previously said the uncertainty surrounding highly advanced AI tools lies in “the unknown unknown,” reflecting broader concerns among policymakers about unpredictable risks.
At the same time, governments have shown interest in testing the technology. Despite ongoing legal tensions between Anthropic and parts of the US Department of Defense over AI usage agreements, federal agencies have reportedly continued evaluating the Mythos system.
The release also comes as Anthropic prepares for a potential public listing, with private valuations approaching $1 trillion. The company’s growing capabilities are seen as central to its appeal among investors in the increasingly competitive artificial intelligence sector.
Jack Clark, co-founder of Anthropic, has previously warned that AI development is accelerating faster than governance frameworks. He has argued that the industry currently lacks adequate mechanisms to slow down or regulate progress if risks become too great, describing the situation as one where technological advancement continues without an effective “brake pedal.”
