Anthropic announced that its large language model Claude identified an enzyme system with properties reminiscent of Crispr, the Nobel Prize-winning gene-editing tool. According to the company, about 950 Claude agents working simultaneously found the genetic sequences in 21.5 hours. The system, called ART (array-associated reverse transcriptases), was found in jumbo phages and contains repeat DNA sequences that resemble Crispr.
While some researchers praised the speed of the discovery, others were more reserved. Fyodor Urnov, a gene-editing expert at UC Berkeley, complimented Anthropic for telling the world, but Stanford professor Le Cong noted that the experiments are still in the queue while the PR is already live. Seth Shipman of the Gladstone Institutes said the novel thing is how they found it, not what it is, and cautioned against attributing the discovery entirely to an AI model.
The same reverse transcriptase was previously identified in a 2021 paper by Jason Gill of Texas A&M, but Claude's contribution was spotting the repeats around it that hint at a Crispr-like system. Still, experts say it remains unknown whether ART can actually cut, copy, and paste DNA in a useful way. Anthropic itself acknowledged it doesn't yet understand what the system does, and the technical report has not been peer-reviewed.
Shipman called the speed impressive, since mining genome databases manually can take months. But he stressed that human scientists must prove the finding in the lab. The opacity of the model's training data also raises reproducibility concerns, and at least one researcher suspected Claude was trained on his unpublished work, which Anthropic denied. For now, the discovery is a promising lead, not a proven tool.