Anthropic announced last week that AI agents in its biology lab had discovered new enzymes with gene editing potential, but a scientist who says he identified the same molecules years earlier is disputing the claim.
The New York Times reports that Mario Rodríguez Mestre, a computational biologist at the University of Copenhagen, said this weekend that he and his colleagues had been studying the enzymes for four years, well before Anthropic’s announcement. Anthropic calls the enzymes and related molecules “ARTs,” short for array-associated reverse transcriptases, and says its AI system took “the first steps of a biological discovery on its own.”
Breitbart News previously reported on the supposed AI-powered discovery:
Anthropic is framing the discovery as an early test of whether Claude can actually be useful to scientists, and the timing is notable: it comes as the company prepares to go public in a potential trillion-dollar IPO. Human involvement, Anthropic says, was limited to writing an initial prompt and running the physical lab work. The AI did the searching and analysis itself.
The process started with a search through a massive database of DNA sequences. Nearly 950 Claude agents worked through the data over 21 hours, processing 210 million tokens. One of those agents noticed an unusual repeating pattern in the dataset and flagged it for human review. Further analysis and lab testing followed, revealing what Anthropic calls “a previously uncharacterized enzyme system found in bacteriophages,” the viruses that infect bacteria.
Mestre says he first came across the enzymes in 2022 while searching for new reverse transcriptases in jumbo phages, viruses that infect bacteria. He later joined the University of Copenhagen to study the enzymes in jumbo phages, which he and his colleagues nicknamed “jumbotrons.” He had briefly consulted for ReNegade Therapeutics, which filed a 2023 patent covering several reverse transcriptases, including ones Mestre found; he is listed as a co-inventor. His and his colleagues’ findings on the jumbotron enzymes have not yet been published.
“So, for me, the most important question is not ‘Were ARTs already known?’ They were,” Mestre said.
For the past three years, Mestre’s team has regularly used Anthropic’s Claude AI models for coding, drafting manuscripts and other tasks. He shared key findings about the enzymes with Anthropic through Claude, including a draft of his dissertation, a manuscript on jumbotrons, and additional details uploaded during what he considered routine administrative tasks. The Times reviewed Slack messages, figures and other materials documenting his research and his conversations with Claude over several years.
Anthropic said Claude’s key contribution was identifying RNA molecules and a protein tied to reverse transcriptase within the system. Mestre says he and his colleagues made that discovery more than a year ago.
In its research, Anthropic had instructed AI agents to search databases containing billions of gene sequences drawn from ocean and soil habitats, most of them from microbes and viruses whose functions are still largely unknown. The exact prompts used were not disclosed. Anthropic scientists gave the AI five examples of reverse transcriptases, enzymes that read RNA sequences and produce matching DNA, and are used by bacteria to sense viruses, while some viruses use them as weapons against other viruses. The agents then wrote their own software to search the databases and analyze candidates, identifying possible reverse transcriptase genes in jumbo phages near arrays of RNA genes.
Anthropic scientists also said the AI independently retrieved raw data from a 2022 experiment recording viral RNA molecules produced by a jumbo-phage-infected microbe, and that the AI reported the microbes had created high levels of RNA molecules in ARTs, supporting a hypothesis about the enzymes’ role in jumbo phages. Mestre said he and his colleagues had analyzed that same 2022 data about a year ago, work that helped them identify RNA genes associated with jumbotrons, and that they had shared that unpublished analysis with Claude.
“My concern is that this information was used to train future versions of the models,” Mestre said. “I think it’s important to raise this possibility.” He also noted that AI companies have previously faced accusations of training on artists’ and writers’ work without permission.
Anthropic denied it had prior knowledge of Mestre’s research and denied training its models on user transcripts. “We are not aware of any previously published work describing the ART system we recently found. Claude was also not trained on any user transcripts, and our molecular biology team has no such access, either,” the company said in a statement Sunday.
Seth Shipman, a bioengineer at the Gladstone Institutes and the University of California, San Francisco, said he was familiar with Mestre’s work and that many researchers in the field already know about these enzyme systems. Shipman said he and his colleagues also rely heavily on Anthropic’s tools and share their own work with the company. Jumbotrons are rare, and their RNA genes are hard to tell apart from random genetic sequences.
A similar dispute played out this month with OpenAI, which claimed to have solved a longstanding math puzzle. NYU mathematician Tristan Buckmaster said he and a colleague had been using OpenAI’s models while working on their own solution to the same problem. OpenAI denied training its models on Buckmaster’s conversations.
Anthropic is expected to go public, at a potential valuation of up to $2 trillion. On Wednesday, chief executive Dario Amodei gave a speech to the United Nations Security Council touting the company’s biological research applications. “We believe this type of application will accelerate rapidly in the years ahead, and that we can make profound progress in the near future on diseases that have plagued humanity for millennia,” he said.
Mestre began using Claude Science, a new Anthropic tool for researchers, last month. After Anthropic’s announcement last week, he said he is winding down projects that use Claude and shifting to other AI models.
Breitbart News social media director Wynton Hall recently unmasked the leftist “effective altruism” philosophy that guides both AI kingpins and the supposed nonprofits would would contribute to AI regulation. Hall has written his New York Times bestseller Code Red: The Left, the Right, China, and the Race to Control AI to serve as the definitive guide on how the MAGA movement can create positions on AI that benefit humanity without handing control of our nation to the leftists of Silicon Valley or allowing the Chinese to take over the world.
Read more at the New York Times here.
Lucas Nolan is a reporter for Breitbart News covering issues of AI, free speech, and online censorship.