Anthropic’s Claude Fable 5 Faces Criticism for Over-caution

WHO:

NBC News - Anthropic is the company releasing the AI model Claude Fable 5.

Fast Company - Derya Unutmazon X is a scientist who commented on the AI model's restrictions.

Fast Company - Bojan Tunguz is a founder and developer who criticized the model on social media.

Fast Company - Pliny the Liberator is an AI researcher who claimed to bypass Fable 5's filters.

WHAT:

NBC News - Anthropic launched Claude Fable 5, its most powerful public AI model.

Fast Company - Users reported that the safety system of Claude Fable 5 was blocking benign or legitimate prompts.

Fast Company - Fable 5 is the first public model derived from Anthropic’s Mythos family.

Fast Company - Anthropic stated that about 0.05% of queries are routed to Claude Opus 4.8 due to safety concerns.

Fast Company - Anthropic acknowledged that it made the wrong tradeoff in designing the classifiers for the model.

Fast Company - Anthropic is working to refine the classifiers to reduce false positives in query responses.

WHERE:

Fast Company - The AI model is being discussed and critiqued on social media platforms.

WHEN:

Fast Company - Claude Fable 5 was launched on a Tuesday.

Fast Company - Reports of false positives began within two days of the launch.

WHY:

Fast Company - Anthropic grouped cybersecurity with other high-risk domains due to concerns raised during the training of the Mythos family.

Fast Company - The company aimed to balance accuracy with transparency in its safety measures.

HOW:

Fast Company - Anthropic's safety system routes flagged prompts to a less capable model, Claude Opus 4.8.

Fast Company - Pliny the Liberator claimed to have used a multi-agent approach and various techniques to bypass Fable 5's filters.

(UPDATED June 11, 2026 23:28 UTC)

Anthropic has launched its most powerful public AI model, Claude Fable 5, which is the first public model derived from the Mythos family. However, users have reported that the model's safety system is blocking benign prompts, leading to concerns that the company may have made the wrong tradeoff in designing its classifiers. As discussions and critiques unfold on social media, Anthropic is actively working to refine these classifiers to reduce false positives in query responses.