Version of AI tool too powerful for public released to public

Anthropic has released Claude Fable 5, a version of its highly advanced AI model Claude Mythos, to the public after initially restricting it due to concerns over its capabilities. The company stated that Fable’s abilities surpass any previously released model, though it will include safeguards and user limitations to mitigate risks. Anthropic warned that releasing such a powerful model carries inherent dangers, despite its precautions. Claude Mythos was first unveiled in April to a select group of organizations, including government and financial leaders, who expressed concerns about potential financial security risks. Anthropic emphasized that the model’s intelligence posed dangers, though some questioned whether the apprehension was overstated marketing. The company now plans to offer Mythos 5, without cybersecurity or biology restrictions, to roughly 150 preview groups, primarily cyberdefenders and infrastructure providers. Access to Mythos 5 will eventually expand through a broader trusted access program, Anthropic announced. Both Fable and Mythos can operate autonomously for extended periods, a capability not seen in prior Claude models. The company’s co-founder, Jack Clark, previously suggested the AI industry lacks controls to slow rapid advancements, comparing it to an unchecked gas pedal without brakes. Anthropic, valued near $1 trillion privately, is preparing for an upcoming public offering. The company’s decision to release Fable 5 reflects a balance between innovation and risk management, though it acknowledges the challenges of deploying such advanced technology.