Anthropic's Claude AI Discovers Novel Enzyme System With CRISPR-Like Properties

Generative artificial intelligence has crossed the threshold from assisting biological research to actively driving it. In a 21-hour autonomous search, Anthropic's Claude AI model identified a previously uncharacterized enzyme system in bacteriophage DNA that shares structural similarities with the CRISPR gene-editing tool.[1][2]

The discovery, announced on September 23, 2026, marks the first major finding from Anthropic's newly established life sciences laboratory in the San Francisco Bay Area. Rather than merely summarizing papers or formatting code, a swarm of approximately 950 Claude-powered software agents combed through public genomic databases, processing 210 million tokens of biological data to spot a pattern that human molecular biologists had overlooked.[3][4]

The newly identified system, dubbed "array-associated reverse transcriptases" or ART, consists of three distinct components. It features a reverse transcriptase—an enzyme that copies RNA into DNA—paired with a partner gene of unknown function and a long sequence of repeating DNA.[2][4]

It is this repeating sequence that caught the attention of both the AI and human researchers. The layout closely mirrors the architecture of CRISPR, the natural bacterial defense system that revolutionized modern medicine by allowing scientists to precisely cut and edit genetic material. "Because the identified array is 'CRISPR-like', some have leaped to conclude Anthropic discovered the 'next CRISPR'," noted a researcher speaking to Al Jazeera, though they cautioned that the system's actual function remains unproven.[1][3][4]

To verify the AI's theoretical finding, Anthropic's human scientists synthesized the ART system and tested it in their physical laboratory. The initial experiments confirmed that the DNA array is actively expressed as short, distinct pieces of RNA. In published data from a Staphylococcus phage, these RNAs constituted up to 8% of the virus's total RNA just 15 minutes after infection.[2][3]

Despite the successful physical validation of the structure, the exact biological mechanism of ART remains a mystery. Anthropic concedes that it has not yet demonstrated whether the enzyme is active or if the system can actually be programmed to cut, copy, or paste DNA in the way CRISPR does. "We don't yet understand what this system does," the company stated in its technical report.[2][4]

The broader significance lies in the method of discovery. The Claude agents autonomously filtered 1.9 billion protein clusters down to roughly 200,000 reverse transcriptases, eventually narrowing the field to 20 highly promising candidates. One agent specifically flagged the distinctive repeat pattern sitting adjacent to the enzyme gene—a needle in a genomic haystack found in less than a day.[4]

Anthropic CEO Dario Amodei expressed optimism about the trajectory of AI-driven biology. "I'm genuinely rooting for all of the frontier labs to seriously get into biological discovery, and I'm excited about what comes out of it," Amodei said, suggesting that the molecular machine could eventually represent a new gene-editing mechanism.[1]

Independent experts have welcomed the development, though with measured expectations. Feng Zhang, a professor at MIT and the Broad Institute who pioneered CRISPR genome editing, reviewed the un-peer-reviewed preprint and called it "an exciting example of how AI agents can contribute to biological discovery."[3]

The next phase of research will require extensive wet-lab experimentation to determine if ART possesses therapeutic utility. While Anthropic has opened the door to AI-led discovery, the rigorous process of proving a biological tool's safety and efficacy remains firmly in human hands.[2][4]