Anthropic formed its life-sciences research group and Bay Area lab in spring 2026 to test whether general-purpose AI could speed the hunt for unfamiliar biological machinery. Its pitch is not that AI replaces scientists: human researchers set the initial task and performed the laboratory work, while Claude agents searched genomic data and generated candidates for review.¹

The first result came after roughly 950 Claude agents spent 21 hours examining a huge DNA database, using 210 million tokens. The system gathered more than 200,000 reverse transcriptases, narrowed them to 3,500 candidate systems and then singled out an unusual sequence pattern beside one enzyme. Anthropic says the resulting system, dubbed array-associated reverse transcriptases, or ART, was found in bacteriophages and combines an RT, a neighboring gene and repeating non-coding DNA.¹

That repeat array is what invited the CRISPR comparison. Claude’s own analysis called the adjacent DNA “spectacular,” asking whether it was a “CRISPR-like … repeat array?” Anthropic says ART has features seen together in only a small number of programmable DNA-processing systems, but stresses that its primary function remains under investigation.¹

Dario Amodei struck the same cautious note while framing the announcement ambitiously. He said the molecular machine “could represent a new gene editing mechanism,” but added that its precise function, biotech utility “(if any),” and ultimate significance were not yet clear.Dario Amodei (DarioAmodei)³

Outside coverage argues that caveat is the central fact. The announcement is Anthropic’s first from its new wet lab and arrives as the company seeks to draw scientists and expand work including drug discovery. Yet a system that resembles CRISPR is not necessarily a tool like CRISPR: whether ART has any practical application, much less transformative potential, remains unknown.²

For now, the durable claim is narrower but still notable: Claude identified a biological pattern researchers had not previously characterized. Turning that pattern into an explanation—or a technology—will depend on the human experiments still ahead.¹