Anthropic Claims Claude Discovered a New CRISPR Enzyme; Researchers Call It Routine Genome Mining
Anthropic recently touted a breakthrough: its AI model Claude independently discovered a novel enzyme system that could expand CRISPR gene-editing capabilities. But CRISPR researchers quickly pushed back, calling the finding standard genome mining rather than a true discovery.
The core dispute centers on what qualifies as an AI-driven scientific breakthrough. Anthropic framed Claude’s output as a novel finding. Experts argue the system merely identified a known enzyme family through conventional database searches, a task performed routinely by human researchers.
What Claude Actually Did
Anthropic provided Claude with prompts about CRISPR-associated systems. The model returned a candidate enzyme from the Cas9 family, specifically a type II-C variant found in Methanomethylophilus alvus.
Key points from the controversy:
- Claude’s output matched an enzyme previously documented in scientific literature. Researchers note this is not a new discovery but a rediscovery of existing data.
- The AI relied on publicly available genomic databases. This is standard practice for genome mining, which humans perform using BLAST and other tools.
- Anthropic claimed the model “reasoned” to find the enzyme. Critics say the model simply retrieved information available in its training data.
The Researchers’ Response
“This is not a discovery. It is routine genome mining that any graduate student could do with basic bioinformatics tools.” — CRISPR researcher commenting on the claim.
Leading gene-editing experts emphasize that AI tools are useful for data sifting, but true discovery requires experimental validation. Anthropic did not conduct lab tests to confirm the enzyme’s function or novelty.
Why This Matters
The incident highlights a growing tension in AI and science communication:
- Overclaiming AI capabilities can mislead the public and investors. Anthropic’s press release suggested a level of ingenuity that the AI does not possess.
- Real AI contributions in biology include predicting protein folding (AlphaFold) or designing new proteins. This case is far simpler.
- The definition of “discovery” is contested. Finding a known enzyme in a database is not the same as uncovering a new natural system.
The Broader Context
Anthropic’s Claude is a large language model trained on vast scientific literature. It can generate plausible hypotheses but lacks the creativity and experimental rigor of human researchers. The enzyme it “discovered” was first described in 2019 by a research team.
CRISPR researchers have been using computational tools for years to mine genomes for new Cas variants. This process is standard and often leads to incremental advances, not headline-worthy breakthroughs.
What Should Have Happened
For a claim of AI discovery to hold weight, experts recommend:
- Validating the output with wet-lab experiments.
- Comparing against known databases to establish novelty.
- Providing full transparency about the AI’s reasoning and training data.
Anthropic did none of these before making the announcement.
The Takeaway
AI models like Claude are powerful assistants but not autonomous scientists. The enzyme system in question is not new. The real lesson is that AI hype can outpace reality, especially in fields where public understanding is limited.
This episode underscores the need for careful communication from AI companies. Scientific credibility depends on humility and rigorous validation, not press releases.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.