Anthropic Apologizes for Undisclosed Claude Fable AI Guardrails
AI developer Anthropic has issued an apology for discreetly implementing guardrails on its new AI model, Claude Fable 5. These hidden restrictions reportedly undermined researchers and competing developers utilizing the system. The company stated it is reversing its approach and will enhance transparency regarding when these restrictions become active, potentially leading to more queries being refused by Fable.
Anthropic has apologized for stealthily incorporating hidden guardrails into its new AI model, Claude Fable 5. These undisclosed restrictions were identified as undermining both researchers and rival companies who were using the model to develop their own competing systems.
The company announced it is reversing this policy. Anthropic committed to greater transparency regarding the activation of these restrictions, acknowledging that this change might result in Fable refusing more queries.
Claude Fable 5 represents the first widely available model within Anthropic's Mythos class of AI systems. The company had previously expressed concerns over the safety of this group of AI systems for public release. Anthropic noted that it had introduced safeguards with Fable to prevent the model from responding to certain "high-risk" queries.
(Source: The Verge)
Advertisement
AdSense slot • inline
