Anthropic's 'AI discovery' claim draws biologist backlash as MIT Tech Review questions the bar

Last Wednesday, Anthropic announced that earlier this year it had launched a molecular biology lab. There, Claude agents read and conjecture about hard biology problems while human scientists run experiments on what they report. According to MIT Technology Review, the company said the lab had made its first discovery. The claim did not land well.

What the system of 950 agents actually found, after 21 hours of work, was not a brand-new DNA sequence. The agents flagged a repeating pattern surrounding a known enzyme, a pattern Anthropic said had not been catalogued before. MIT Technology Review, in a piece published on 28 September and republished in The Algorithm newsletter, noted that Anthropic's announcement describes the pattern as “reminiscent” of what led to CRISPR, the gene-editing technology that “has already transformed science and medicine.” That framing is where the trouble starts.

Biologists say finding the pattern was the easy part

A viral post from a biologist, subsequently 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.” MIT Technology Review reported that the agents helped with some laboratory grunt work, but a discovery it is not, according to that critique.

Muddying things further, Mario Rodríguez Mestre, a biologist at the University of Copenhagen, said over the weekend that his team had already discovered this particular pattern. The New York Times reported the claim, according to MIT Technology Review. Mestre, who regularly chatted with Claude in his work, wondered whether Anthropic's team had learned from his conversations. Anthropic denies this, but Mestre says he is stopping all use of Claude anyway. The dispute matters beyond one company's press release. MIT Technology Review argued that AI companies are not presenting their systems simply as tools scientists can use, like microscopes or supercomputers, but insisting the systems are making discoveries themselves. To some biologists, that is incompatible with how science actually works, where new knowledge more typically emerges from collaboration and an ever-growing arsenal of tools.

The math problem next door

The pattern is not unique to biology. MIT Technology Review noted that earlier this month OpenAI said its own team agents had cracked a million-dollar problem in mathematics. A couple of weeks later, nearly every AI skeptic in the writer's feed was sharing an article asking whether it was the math problem that really mattered. The piece did not argue OpenAI's solution was wrong, only that the particular result may not be the one mathematicians care most about. Throw in an accusation by a mathematician that the models may have used some of his work without credit, and people are left thinking either OpenAI cheated or the solution was not important anyway.

Lucas Harrington, the biologist who wrote the post critiquing Anthropic's announcement, closed with a suggestion. AI companies, he said, 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.” MIT Technology Review noted that raising the bar might be the last thing on the minds of OpenAI's Sam Altman and Anthropic's Dario Amodei as they race to one-up each other.

Real results, different questions

Elsewhere the science is less contested. A paper submitted to arXiv on 15 September and posted to the preprint server reports the first direct detection of auroral radio emission from an exoplanet, the giant planet β Pictoris b, using the MeerKAT array. The authors, Kevin N. Ortiz Ceballos, Edo Berger and Yvette Cendes, report rapid, recurring, highly circularly polarized bursts as well as persistent emission at frequencies of 0.85 to 3.5 GHz. They identify the emission as electron cyclotron maser radiation, which implies a magnetic field of at least 1.25 kG at the planet, the first such direct field strength measurement for an exoplanet. The paper is 32 pages with nine figures and three tables.

In medicine, a study published in Cell Reports describes a molecular mechanism that may explain why lost weight so often returns. Researchers at the Harrington Discovery Institute at University Hospitals and Case Western Reserve University found that obesity altered the activity of fat cells in a way that kept asprosin, a hormone released by fat tissue that signals the brain to stimulate appetite, elevated after weight loss. The work centers on TGF-β1, an inflammatory signaling molecule; brief exposure to it created a lasting epigenetic change that remained for weeks after TGF-β1 levels returned to normal. Even when mice shed all their excess weight, their fat cells continued producing more asprosin.

“It was like flipping a light switch that stays on even after you remove your finger,” said Atul Chopra, MD, PhD, senior author of the study and associate director of the Harrington Rare Disease Program at Harrington Discovery Institute at UH. “Even after mice lost all the excess weight and TGF-β1 returned to normal, this switch in their fat cells remained flipped, keeping asprosin and appetite elevated. This gives us a molecular explanation for why GLP-1 drugs, which suppress appetite while you take them, cannot fix the underlying biological memory that drives hunger back up once treatment ends.”

Scitech Daily, which reported the findings on 28 September, noted that evidence from publicly available human datasets supported the proposed mechanism, although researchers must still confirm directly that it operates the same way in people. A separate laboratory led by Seth J. Field, MD, PhD, confirmed that the cellular change persisted.

What to watch

MIT Technology Review will publish its 2026 list of Climate Tech Companies to Watch on 6 October, a package covering energy storage, nuclear power and transportation. The newsletter also flagged that OpenAI has scrapped a new AI model, GPT-6.1 Astra, over safety concerns, with the company saying it “didn't quite meet the bar,” according to BBC. Anthropic's IPO filing warns that AI could threaten humanity, the FT reported.

For now, the disagreement over Anthropic's enzyme pattern is unresolved. The company denies that its system learned from Mestre's Claude conversations. Mestre says he has stopped using Claude. And the broader question, whether an AI flagging a pattern in a mass of biological data counts as a discovery, remains open.