Anthropic says its Claude AI model has autonomously discovered a previously unknown enzyme system in bacterial DNA that shares key features with CRISPR. The finding is drawing attention as a potential path toward new gene-editing tools, though scientists remain divided on how significant it really is.

950 AI Agents, 21 Hours of Searching

The discovery came out of Anthropic's newly established biology research lab in San Francisco. Working under direction from human researchers, Claude deployed roughly 950 agents that spent 21 hours combing through a massive database of DNA sequences, consuming about 210 million tokens in the process.

While hunting for new examples of reverse transcriptase enzymes, the agents spotted an unusual pattern: a set of evenly spaced repeating DNA sequences sitting right next to a reverse transcriptase gene. Because that repeat structure resembled the arrays found in CRISPR systems, Anthropic is calling the newly identified system "array-associated reverse transcriptases," or ART.

A "Programmable" Structure Reminiscent of CRISPR

The ART system consists of three parts: the reverse transcriptase enzyme itself, an associated partner gene, and the evenly spaced array of DNA repeats. Early experiments show that this repeat region produces several distinct short RNAs, hinting that ART could turn out to be "programmable" in a way similar to CRISPR.

CRISPR itself originated as a bacterial immune mechanism before being adapted into a tool for cutting and editing DNA, a breakthrough that has already opened the door to treatments for diseases such as sickle cell disease and cancer. The scientists who pioneered turning naturally occurring CRISPR into that editing tool, Emmanuelle Charpentier and Jennifer Doudna, won the Nobel Prize in Chemistry in 2020 for the work.

Excitement Mixed With Caution

Anthropic CEO Dario Amodei said on social media that the company suspects this "molecular machine" could represent a new gene-editing mechanism, while stressing that AI-driven science is still at a "very early" stage when it comes to producing medical breakthroughs. Stanley Qi, an associate professor of bioengineering at Stanford University, praised the AI's ability to spot a hard-to-detect biological pattern and pursue it systematically as a research question.

Feng Zhang, a CRISPR pioneer at MIT and the Broad Institute, called it an exciting example of how AI agents can contribute to biological discovery.

Not everyone is convinced the finding is as significant as it sounds. Kevin Blake, a microbiologist at Washington University School of Medicine, noted that CRISPR-like repeat sequences are already common in nature, and that countless bacterial species have yet to be studied and catalogued. He added that there is currently no evidence the system rivals CRISPR as a technology or could be developed into a therapeutic or practical application.

Summary

Anthropic has not yet determined what function the ART system actually serves, and further experiments are underway. The announcement highlights how AI could dramatically speed up scientific discovery, while also raising the question of how cautiously such early findings should be interpreted. As debate continues over AI's risks and potential, how far this kind of biological application can go remains to be seen.