Anthropic says Claude found a novel CRISPR-like enzyme system

By Splitfeed EditorialLast updated Sep 29, 2026, 19:21 UTC

In late September 2026, Anthropic announced that Claude agents operating inside a molecular biology lab the company had launched earlier that year identified a novel enzyme system in bacteriophages with a CRISPR-like structure. The lab's model has Claude agents reading scientific literature and generating hypotheses, while human scientists run the physical experiments that follow. Anthropic framed the finding as a meaningful demonstration of AI-driven discovery, specifically pointing to Claude's identification of an array of RNA molecules and a protein associated with reverse transcriptase as the key contribution.

The announcement almost immediately ran into two distinct lines of challenge. The first is methodological: a scientist quoted by Wired noted that the experiments validating the discovery were still in the queue when Anthropic went public, raising the question of whether an unvalidated hypothesis constitutes a scientific finding at all. The second challenge is one of priority and attribution. A researcher identified in the Anti-AI camp's quoted reactions as Dr. Mestre has stated that he and his colleagues had already made the same discovery about transcription enzymes and had shared that work with the company, suggesting Anthropic may have announced as novel something that was not new to science. That second claim, if substantiated, would shift the story from a debate about AI capability to a question of research ethics and data practices.

The Anti-AI camp has been the most vocal. Accounts including Barbara J. King, Mindy Weisberger, and Hagen Blix have all pointed to the overlap between Anthropic's announced finding and Dr. Mestre's prior work, with Blix suggesting the result may have emerged from training on user-submitted data. Juspreet Singh Sandhu raised the phrase "academic plagiarism" directly, while Ben Eltham and Jeff Blattner used the moment to question Anthropic's broader credibility given the company's own statements about existential risk from AI. No Pro-AI or Middle Ground camp reactions have been published yet. In a story like this, the Pro-AI camp would typically argue that AI-assisted hypothesis generation represents a genuine acceleration of the scientific process regardless of how credit is ultimately assigned, and the Middle Ground camp would typically call for clearer standards about what constitutes an AI discovery before either celebrating or condemning the result.

The argument is unlikely to be settled until the experimental validation is complete and the question of Dr. Mestre's prior work is addressed more formally, whether through peer review, institutional review, or a direct response from Anthropic about how and when his findings entered their pipeline.