# Anthropic’s AI biology finding raises questions about scientific discovery

> Anthropic says 950 Claude agents identified a previously uncatalogued pattern near a known enzyme. Biologists disagree over whether finding that pattern counts as a scientific discovery.

Oossa · 2026-09-28 · https://oossa.com/en/anthropic-s-ai-biology-finding-raises-questions-about-scientific-discovery

Anthropic says its AI system has made its first discovery in biology. The company’s lab uses Claude agents to study difficult biology questions, while human scientists carry out experiments. In this case, the agents searched biological data and flagged a repeating pattern around a known enzyme.

The company says the pattern had not been catalogued before. But the agents did not find a new DNA sequence or establish what the pattern does. That distinction matters: spotting something unusual can be a useful step in research, but it may take experiments and further study to show why it matters.

## What did the agents find?

Anthropic says 950 agents identified the pattern in 21 hours, after narrowing down 200,000 candidates. Its announcement compared the pattern to discoveries that helped lead to CRISPR, a gene-editing technology. Some biologists objected to that framing. Lucas Harrington, a biologist who criticized the announcement, said identifying an unusual cluster is often easier than working out what it does.

There is also a dispute about whether the pattern was new. Mario Rodríguez Mestre, a biologist at the University of Copenhagen, told the New York Times his team had already found it. He wondered whether Anthropic’s researchers had learned about it through his conversations with Claude. Anthropic denies this; Mestre says he will stop using Claude.

## Why the label matters

The debate is partly about what people mean when they say AI made a discovery. The agents may have helped sort a large set of candidates, but humans steered the work and ran the experiments. Calling the result an AI discovery can make a useful contribution sound like a complete scientific breakthrough—or invite criticism if it falls short of that standard.

The article argues that companies should describe what their systems actually did and set a high bar for claims of discovery. Otherwise, genuine progress may be dismissed, while a finding’s importance gets lost in arguments over whether the AI deserves credit.

## The facts

- Anthropic says 950 Claude agents found the pattern after 21 hours of work.
- The agents narrowed 200,000 candidates to a repeating pattern around a known enzyme.
- Anthropic says the pattern had not been catalogued before.
- Biologist Mario Rodríguez Mestre says his team had already found the pattern; Anthropic denies learning about it through his Claude conversations.

## Why it matters

AI tools may help researchers sift through far more data than people can inspect by hand. For the public, the key question is what the system actually contributed—and whether experiments show that a finding changes our understanding of biology.

## Sources & references

1. [When can we say AI made a scientific discovery?](https://www.technologyreview.com/2026/09/28/1145230/when-can-we-say-ai-made-a-scientific-discovery/) – MIT Technology Review, 2026-09-28

Last updated: 2026-09-28