Anthropic's AI biology lab says it made a discovery. Biologists disagree.

Anthropic announced last Wednesday that a molecular biology lab it launched earlier this year had made its first discovery. Inside the lab, Claude agents read and conjecture about hard biology problems, while human scientists design and run experiments on what the agents report.

What the agents actually found

After 21 hours, a system of 950 agents did not present a brand-new DNA sequence. It flagged a repeating pattern surrounding a known enzyme, one that Anthropic said had not been catalogued before. The company's announcement described the pattern as "reminiscent" of the work that led to CRISPR, the gene-editing technology that, in its words, "has already transformed science and medicine."

That framing angered some biologists. A viral post by Lucas Harrington, later endorsed by the chair and CEO of the drugmaker Eli Lilly, argued that "finding a weird cluster of genes and repeats is often the easy part. The hard part, and where the real discoveries come from, is figuring out what the system actually does." The agents helped with laboratory grunt work, he wrote. A discovery it was not.

A claim that had already been made

Over the weekend the biologist Mario Rodríguez Mestre, of the University of Copenhagen, said his team had already found the same pattern, the New York Times reported. Mestre, who regularly chats with Claude in his work, wondered whether Anthropic's researchers had learned from those conversations. Anthropic denies drawing on his exchanges; Mestre says he is stopping all use of Claude anyway.

The dispute is not only about one pattern. AI companies are not presenting their systems as tools scientists can pick up, the way they use a microscope or a supercomputer. They insist the models are making discoveries themselves, an approach some researchers consider incompatible with how science works, where knowledge usually emerges from collaboration and an expanding arsenal of instruments.

Why the framing matters

Narrowing 200 000 candidates down to a handful worth exploring is legitimate scientific work, and the fact that a general-purpose chatbot can do part of it is notable even with humans steering and running the experiments. But once the question becomes whether Claude itself made a discovery, the debate collapses into two answers: breakthrough or bust. That framing also invites moving the goalposts after a real win.

Something similar happened earlier this month, when OpenAI said its agents had cracked a million-dollar mathematics problem. Two weeks later, skeptics were sharing an article asking whether the problem it solved was the one mathematicians actually care about.

Harrington closed his post with a suggestion: AI companies should "set the bar high now, so that when an AI actually discovers a fundamentally new biological mechanism, everyone appreciates how big a deal it is." As OpenAI's Sam Altman and Anthropic's Dario Amodei race to outdo each other, raising that bar may be the last thing on their minds.