Anthropic announced on September 23, 2026 that it had formed a life sciences research group and laboratory to use Claude in fundamental biology research. The company says one of the group’s first campaigns produced an early result: the identification of a novel enzyme system with repeats resembling CRISPR arrays. The claim comes from Anthropic’s own account, so it establishes what the company reports rather than independent confirmation of the discovery.
According to the account, roughly 950 Claude agents searched a large DNA-sequence database for 21 hours, using about 210 million tokens. They reportedly gathered more than 200,000 reverse transcriptases, selected 3,500 candidate systems, and narrowed the set to 20 for human-readable analysis. One candidate led researchers to an uncharacterized system in bacteriophages, which Anthropic calls array-associated reverse transcriptases, or ART.
The proposed system has three reported parts: a reverse transcriptase, a neighboring partner gene, and a long array of evenly spaced DNA repeats. Anthropic says its first experiments found that the array is expressed as distinct short RNAs, suggesting a possible analogy with CRISPR, whose RNA-guided repeats help make some systems programmable. The underlying reverse transcriptase had been identified previously; Anthropic says Claude was the first to notice this particular combination of a repeat array and an accessory protein of unknown function.
The most important limitation is also the clearest one: Anthropic says the primary function of ART is still unknown. The supplied material does not show that the system cuts, copies, pastes, or edits DNA, nor does it establish a therapeutic use. Anthropic characterizes the work as an early finding and says further experiments are underway.
The practical consequence is therefore narrower than “AI discovered a new CRISPR.” The report suggests that AI agents may help scan enormous sequence collections and flag unusual genomic arrangements for scientists to investigate. Whether ART becomes a useful biological tool depends on experiments that establish what it does, how it works, and whether its behavior can be controlled.