Anthropic releases a safeguarded Mythos-class model for broader use (Fable 5)

Anthropic launched Fable 5, a general-purpose model built on the same foundation as its previously restricted Mythos-class systems. The company said new safeguards allow broader deployment by automatically redirecting certain high-risk cyber, biology, chemistry, and model-distillation requests to less capable models. Anthropic described the release as its most capable public model to date, with strong performance across software engineering, research, and knowledge-work tasks. The move is a significant shift from the company's earlier position that Mythos-class systems required substantially tighter access controls while additional safety measures were developed. Early testing of Claude 5 Fable, described as a Mythos-class AI model, suggests a substantial jump in autonomous problem-solving capability. Across a range of projects, the model independently conducted research, delegated tasks to AI agents, wrote and tested code, evaluated results, and worked for up to a dozen hours with minimal human intervention. Examples included building a historically inspired isochrone travel map using thousands of transportation data points and creating sophisticated research-analysis software after generating its own 19-page design document. The experience highlighted both the power and opacity of advanced AI systems. Users can provide ambitious goals and feedback, but much of the reasoning, decision-making, and execution occurs inside a black box. The human role is shifting from directing each step of the process to specifying objectives and evaluating outcomes.