Anthropic's Claude Fable 5 Avoids Basic Biology Questions

Anthropic has recently unveiled Claude Fable 5, an AI model they've described as their most powerful to date, with particular praise for its biological capabilities. However, it appears that this cutting-edge AI struggles with even fundamental biology questions, the kind that a high school student might be expected to answer. Instead of providing an answer, Fable 5 redirects these inquiries to the previous flagship model, Claude Opus 4.8.

This isn't due to a lack of knowledge on Fable 5's part. It's a deliberate design choice by Anthropic. Fable is classified as a public-facing, Mythos-class model. This family of AI models is reportedly so adept at cybersecurity tasks that Anthropic previously deemed them too dangerous for public release.

The decision to restrict Fable 5 from answering basic biology questions raises questions about Anthropic's approach to AI safety and public deployment. While the company emphasizes the advanced capabilities of its Mythos-class models, particularly in areas like cybersecurity, this limitation suggests a cautious strategy in making these powerful tools widely accessible. It seems Anthropic is prioritizing safety and control over full disclosure of capabilities, even for seemingly innocuous topics.

For users expecting Fable 5 to be an all-knowing AI, this limitation might come as a surprise. It highlights the complex balancing act AI developers face: pushing the boundaries of artificial intelligence while managing the potential risks and societal implications. The move also suggests that even the most advanced AI models might have specific guardrails and limitations imposed by their creators, influencing their behavior and the information they are willing to share.