Anthropic opened its Bay Area biology lab this spring, betting that AI could accelerate work in medicine. Its researchers say Claude then searched biological data with roughly 950 agents, using 210 million tokens over 21 hours, before flagging a bacteriophage enzyme system with DNA repeats that resemble CRISPR. The physical experiments, Anthropic says, remained in human hands and were confined to lower biosafety-level work.¹

On Wednesday, the company presented the result as its first laboratory discovery: genes in viruses that produce enzymes acting in what it believes is a new way. Life-sciences chief Eric Kauderer-Abrams said biology and medicine were Anthropic’s chosen route for delivering AI’s benefits. But the technical report has not yet been submitted to a scientific journal.²

Dario Amodei struck a notably cautious note alongside the fanfare, calling it a Claude-led discovery of a molecular machine that could be a new gene-editing mechanism. He said its function, usefulness and ultimate importance were still unclear.Dario Amodei (DarioAmodei)³

That caution has become the fault line. Philip Kranzusch, a Harvard Medical School microbiologist, said the evidence did not yet justify a breakthrough label: “It’s a wrinkle on what we know in the field.”²

Then, over the weekend, computational biologist Mario Rodríguez Mestre said his group had studied the enzymes and associated molecules—dubbed ARTs by Anthropic—for four years. His team has not published those findings, but he said its members had used Anthropic models while coding and drafting work, potentially sharing key details with the company. “The most important question is not ‘Were ARTs already known?’ They were,” Mestre said, arguing the sharper issue is whether Claude reasoned its way to the result or was steered by information originating in his group.⁴

For now, Anthropic has offered a provocative lead, not a settled scientific victory. Independent validation—and answers about provenance—will determine which it becomes.