Anthropic's Biology Lab Says AI Helped Uncover a New Enzyme System From Viral DNA

Anthropic says researchers at its in-house wet lab used its Claude AI models to identify a previously unknown enzyme system in the DNA of bacteriophages, viruses that infect bacteria, with properties that resemble CRISPR gene-editing tools. The company says hundreds of AI agents were directly involved in the analysis behind the finding, drawing attention to how AI is being woven into hands-on scientific research.

950 Agents, 21 Hours, 210 Million Tokens

The discovery came out of a biology lab Anthropic set up in the Bay Area in spring 2026. According to the company, around 950 Claude agents ran for roughly 21 hours, processing about 210 million tokens as they worked through bacteriophage genome data. One of the agents reportedly flagged the pattern during the analysis, essentially noting that it looked like a CRISPR-style repeat array, which became the starting point for the discovery.

The enzyme itself had already been documented in prior research, but Claude identified surrounding genetic sequences that are only seen in "cut, copy, and paste" DNA systems. Anthropic is tentatively calling this system ART, and says it could point toward a new kind of gene-editing tool. The company has not yet determined the system's precise function or practical applications.

Human Scientists Ran the Experiments, With Safety Limits in Place

Anthropic's lab operates at biosafety levels 1 and 2 and does not work with pathogens that infect humans. The company says human scientists chose the research direction and carried out every physical experiment, and that Claude did not operate any lab equipment autonomously. CEO Dario Amodei said Claude "drove the vast majority" of the discovery, while adding that it may eventually be possible for Claude to safely run experiments by controlling lab equipment on its own, with appropriate safeguards, but that Anthropic is not doing that today. Amodei also acknowledged that researchers at Stanford University had previously identified a system that works in a similar way, and said the broader research community still needs to validate the significance of Anthropic's finding.

Part of a Wider Push to Apply AI to Biology

Anthropic is far from alone in applying AI models to fundamental biology research. Stanford University, UC San Francisco (UCSF), and Google's protein-structure work are among the institutions and companies pursuing similar approaches. As generative AI moves beyond summarizing existing literature and starts contributing to experiment design and hypothesis generation, it's a sign that the process of doing science itself may be starting to change.

Summary

Anthropic says a team of AI agents working inside its wet lab helped uncover a previously unknown enzyme system that could function similarly to CRISPR. Human scientists carried out the physical experiments and safety limits remained in place throughout, making this a notable example of AI playing a direct role in the scientific discovery process. Whether the finding holds up, and what it might be used for, will depend on further outside verification and follow-up research.