Anthropic says Claude autonomously discovered a new enzyme system

By Splitfeed EditorialLast updated Sep 26, 2026, 19:17 UTC

On September 23, 2026, Anthropic announced that its AI model Claude had autonomously identified a new enzyme system by searching a large database of DNA sequences. GenomeWeb, reporting a day later, described the find more precisely as a novel reverse transcriptase and DNA repeat array. Anthropic compared the system to the molecular machinery underlying Crispr gene editing, a technology that transformed biology and medicine after its development in the 2010s. The discovery is also the first public result from Anthropic's newly launched wet laboratory, where Claude's computational findings were followed up with physical lab work carried out by human scientists.

The announcement carries weight beyond a single finding because of what it is meant to demonstrate. Anthropic is preparing to go public, and framing Claude as a tool capable of original scientific discovery positions the company in a market where AI applied to drug discovery and genomics has attracted enormous investment. At the same time, the details of how the discovery unfolded are contested terrain. TechCrunch noted that Anthropic has not operated Claude unsupervised in its biology lab, with humans remaining in the loop throughout. That caveat matters because claims of autonomous scientific discovery invite scrutiny over what autonomous actually means when a human sets the initial prompt and others conduct the physical verification.

None of the three camps, Pro-AI, Anti-AI, or Middle Ground, had published reactions as of the available inputs, so what follows reflects what each would typically argue about a story of this kind rather than anything anyone has said. The Pro-AI camp would likely treat the finding as evidence that AI can accelerate scientific research in ways that justify both the investment and the ambition, pointing to the Crispr comparison as a signal of transformative potential. The Anti-AI camp would probably question whether the discovery constitutes genuine autonomous reasoning or sophisticated pattern matching over existing data, and would press on the commercial timing, arguing that a pre-IPO announcement invites hype over rigorous peer review. The Middle Ground camp would most likely acknowledge real promise in AI-assisted genomics while calling for independent validation and clearer definitions of what human oversight actually looked like at each stage.

The most immediate thing to watch is whether Anthropic submits the reverse transcriptase and DNA repeat array findings to a peer-reviewed journal, which would allow independent scientists to evaluate the claim on its merits and separate the scientific result from the investor narrative surrounding it.