Claude Fable is useless for bioinformaticians
A bioinformatician expresses frustration that the AI model Claude frequently flags routine bioinformatics and computational biology requests as potential biosecurity threats, comparing it to being treated as if trying to weaponize viruses or create anthrax, and asks other life science users about their experiences.
Background
- **Claude Fable** is an AI coding assistant developed by Anthropic, designed to write and run code safely. It employs strict safety filters that block certain types of biological, chemical, or medical prompts to prevent misuse (e.g., designing pathogens or toxins).
- **Bioinformatics** and **computational biology** are fields that analyze biological data (e.g., DNA sequences, proteins) using software. Routine work includes tasks like genome assembly, variant calling, or sequence alignment — which can superficially resemble dangerous bio-weapon research.
- The poster is frustrated that Claude Fable's safety system over-flags legitimate bioinformatics requests, treating common research tasks as if they were attempts to create anthrax or weaponize viruses.
- This reflects a broader tension in AI safety: models trained to block harmful biological queries often cannot distinguish between legitimate research and malicious intent, hampering real scientists who need to use these tools for their work.