Anthropic has reported the first result from a molecular biology group it formed in spring 2026.[1] The group asked Claude to search a large DNA-sequence database for interesting reverse transcriptases, enzymes that copy RNA into DNA.[1][2]
Roughly 950 Claude agents worked for 21 hours, using about 210 million tokens.[1][2] They gathered more than 200,000 reverse transcriptases, assessed thousands of candidate systems and produced reports on the most compelling examples for human review.[1][2]
One agent noticed a repeating DNA pattern beside an unusual reverse transcriptase.[1][2] It counted the repeats, compared the arrangement with known systems and searched existing research before reporting the candidate.[1] Anthropic’s scientists then investigated it in the laboratory and named the system array-associated reverse transcriptases, or ART.[1][2]
ART occurs mainly in bacteriophages, viruses that infect bacteria, and combines a reverse transcriptase, a neighbouring partner gene and an array of evenly spaced DNA repeats.[1][2] Anthropic says the layout has properties reminiscent of CRISPR, while its first experiments indicate that the array is expressed as distinct short RNAs.[1][2]
The comparison does not mean Anthropic has discovered another CRISPR tool.[1][2] Researchers do not yet know ART’s biological function, and the technical report is a preprint that has not been peer reviewed.[1][2][3] The underlying reverse transcriptase had also appeared in earlier studies; the reported discovery is the wider system formed by the enzyme, repeat array and partner protein.[1][2]
The result also exposed a reproducibility problem.[2][3] Anthropic repeated the original search ten times, but none of those runs examined the relevant upstream DNA and all missed the repeat array.[2][3] When models were given the relevant sequence directly they usually recognised the pattern, suggesting that navigating a vast search space remains a major part of the challenge.[2][3]
Human researchers still performed all physical laboratory work.[1][2] The useful division of labour was AI at computational scale, followed by expert review and controlled experiments.[1][2] That makes this less a story about replacing biologists than about giving them a much larger search team for the earliest stages of discovery.[1][2]
Supporting sources
[1] Anthropic: Claude discovers a novel enzyme system with CRISPR-like repeats
[2] Anthropic technical report: Autonomous biological discovery by AI agents
[3] TNW: Anthropic says Claude found a new enzyme system with CRISPR-like repeats
Source: Anthropic