AI Enzyme Discovery: Claude Finds a CRISPR-Like System in Phages

An AI enzyme discovery announced by Anthropic this week has caught the attention of biologists. The company says its Claude agents found a previously uncharacterised enzyme system with features that resemble CRISPR, the tool behind modern gene editing.

The system is found mainly in bacteriophages, the viruses that infect bacteria. Anthropic’s researchers have named it array-associated reverse transcriptases, or ART.

How the AI enzyme discovery happened

Anthropic formed a life sciences research group in the spring of 2026. It runs a laboratory in the Bay Area. According to the company, human scientists gave Claude a single high-level prompt and did the lab work. The agents did the searching.

Roughly 950 agents worked for 21 hours and used about 210 million tokens. They gathered more than 200,000 reverse transcriptases, enzymes that copy RNA into DNA. They flagged 3,500 new candidate systems and narrowed those to 20 for detailed reports.

One agent noticed a long, evenly spaced array of repeated DNA next to an odd-looking enzyme gene. The underlying enzyme had been seen before. The company says Claude appears to be the first to spot the array and a partner protein of unknown function.

What is known, and what is not

ART has three parts: the enzyme, a partner gene and the repeat array. Early lab tests suggest the array is expressed as a set of distinct short RNAs. That hints at a programmable mechanism, as in CRISPR.

But nobody yet knows what ART does. The findings are in a pre-print, not a peer-reviewed paper. Anthropic stresses that further experiments are under way. The lab works only at biosafety levels 1 and 2 and does not handle human pathogens.

Feng Zhang, a CRISPR pioneer at MIT and the Broad Institute, reviewed the pre-print. He called the find “genuinely intriguing” and said it “merits further investigation”.

Why sceptics should watch closely

The claim is striking, and it comes from a company with an interest in promoting its product. Independent replication is the proper test. Still, the method is open to scrutiny, and the pre-print is public.

The news lands in the same week AI leaders warned the UN Security Council about AI risks. Promise and peril are arriving together. Read Anthropic’s announcement and pre-print.

Fact basis: Anthropic announcement (23 September 2026) and pre-print; reporting by The Next Web, Interesting Engineering and Investing.com.