Anthropic chief executive Dario Amodei announced on X that his company’s Claude model discovered a new molecular machine that could potentially represent a novel gene-editing mechanism, sparking fierce debate across the microbiology community. While Anthropic frames the finding as a breakthrough for artificial intelligence in life sciences, researchers at the Institut Pasteur and independent laboratories have questioned the company's communication tactics, data sovereignty, and whether the model actually uncovered anything entirely new.
The announcement immediately divided genome-editing specialists, who questioned the true novelty of the discovery and whether the AI system inadvertently processed proprietary data belonging to working scientists. Lexpress.fr reported on the escalating friction between artificial intelligence developers and structural biologists following the unexpected claim.
Institut Pasteur Experts Question Genome Editing Claims
Aude Bernheim, a bacterial immunity and genome editing specialist at the Institut Pasteur, urged caution regarding the company’s sweeping claims. She stated that the discovery cannot currently serve as a genome-editing tool and noted that nothing in the provided data allows researchers to affirm otherwise. Bernheim pointed out that identifying protein families with unknown functions from vast bacterial genome databases—a process known as genome mining—is relatively straightforward, whereas determining their biological roles and practical utility remains the real challenge.
Ernest Mordret, a genetic AI research expert at the Institut Pasteur, acknowledged that using AI agents to read DNA and spot repeating nucleotide sequences is a novel approach. However, Mordret emphasized that the process relies heavily on human intervention and co-construction rather than fully autonomous discovery. Bernheim added that Anthropic has not yet demonstrated that the identified molecules produce any specific biochemical activity, nor is there a confirmed structural link to the CRISPR systems made famous by Nobel laureates Emmanuelle Charpentier and Jennifer Doudna.
The controversy deepened when researchers noted similarities between the protein family identified by Anthropic and “retrons”—bacterial defense mechanisms that produce a mix of DNA and RNA to ward off viruses. That overlap has led the scientific community to question whether Claude reused existing data from independent researchers, a concern Mordret noted is currently impossible to verify.
Data Sovereignty Concerns Grow Across Biotechnology Laboratories
The dispute over Anthropic’s findings echoes a similar incident in mathematics earlier in the month, when OpenAI’s models tackled the Navier-Stokes equation while independent mathematicians were using the same platform for closely related work. Mordret noted that these parallel developments have left researchers on edge regarding data privacy and intellectual property. At the Institut Pasteur, teams are actively evaluating data sovereignty risks and working internally to design secure tools for ongoing genetic research.
Despite these commercial and methodological tensions, Mordret highlighted that Anthropic has made aggressive strategic hires, including John Jumper, the lead author of AlphaFold who shared in the 2022 Nobel Prize for protein structure prediction. Bernheim warned that while the AI sector’s tendency to exaggerate biological breakthroughs risks alienating the public and creating societal resistance, disciplined and regulated integration of artificial intelligence will remain essential for driving future biological discoveries.
