Anthropic Disputes Fable 5 AI Jailbreak

Anthropic disputed claims that Claude Fable 5 (a powerful AI model with safety restrictions) was jailbroken, which is the process of tricking an AI into bypassing its safety restrictions. A security researcher claimed to have circumvented the model's safeguards using sophisticated multi-agent prompting methods (techniques that chain multiple AI requests together), but Anthropic argued the approach only caused conversational refusals rather than defeating core safety systems, and that independent classifier systems (separate AI models that filter dangerous outputs) still prevented genuinely harmful content.