Claude’s Enzyme Find Draws Praise From Anthropic—and Pushback From CRISPR Researchers

Anthropic says its AI model Claude found a previously unknown enzyme system in DNA databases while doing most of the analysis independently. Humans set the research question and handled the lab experiments. The system resembles CRISPR, the gene-editing tool, but its function is still unknown.

According to the report, about 950 AI agents spent 21 hours reviewing more than 200,000 reverse transcriptases, which are enzymes that convert RNA into DNA. Next to one unusual enzyme, the agents found repeating DNA sequences. Anthropic dubbed the system “ART,” and the report says it appears mostly in bacteriophages, which are viruses that infect bacteria.

Lucas Harrington, who earned his PhD under Nobel laureate Jennifer Doudna and co-founded Mammoth Biosciences with her, was less impressed by Anthropic’s framing. Harrington said genome mining has been used for decades and that similar systems have been known since 2008. His core objection: identifying a system is not the same as showing what it does.

The key takeaway is not that AI has fully automated biological discovery, but that AI may accelerate parts of database-driven research. Anthropic’s claim is notable because Claude handled much of the computational analysis, yet the scientific value still depends on lab validation and functional understanding. Until ART’s role is demonstrated, the finding remains an early lead rather than a completed breakthrough.