ThinkFacility
  1. Home
  2. News
  3. Anthropic

Claude found a new enzyme system with CRISPR-like repeats, but Anthropic doesn't know what it does

Roughly 950 agents searched for 21 hours, and one of them stopped at a run of repeats next to an enzyme scientists had already catalogued.

On September 23, 2026, Anthropic said Claude had found a family of enzymes in viral DNA that nobody had described before. It's the first result from a biology lab Anthropic set up this spring, and the company announced the lab and the finding in the same post.

The system sits in bacteriophages. Beside the enzyme's gene there's a long run of evenly spaced DNA repeats, laid out much like the repeats CRISPR was first noticed for. Anthropic calls it ART, short for array-associated reverse transcriptases.

Agents
Roughly 950
Search time
21 hours
Tokens
210 million
Enzymes gathered
Over 200,000 RTs
Next
Further experiments are underway to determine how ART works

What Claude did on its own

Anthropic's scientists gave Claude one prompt: search a huge DNA database for new examples of reverse transcriptases. They say their involvement "was limited to the initial prompt and the lab work". The agents gathered the enzymes, picked out 3,500 candidate systems and wrote reports on the 20 they judged most compelling.

Feng Zhang in a tuxedo holds open a wooden case with a bronze medal beside a woman in a black dress, with American flags behind them
NSF Director France Córdova presents Feng Zhang with the 2014 Alan T. Waterman Award. Photo: National Science Foundation, Public domain, via Wikimedia Commons

One agent was reading raw sequence next to an odd-looking enzyme when it hit the repeats. Anthropic printed what it wrote.

I can see by eye a tandem repeat array

Claude discovers a novel enzyme system \ Anthropic

It counted the repeats, measured their spacing and searched the literature for any earlier report of the pattern. Then it filed a report for human review. The enzyme itself had turned up in earlier studies of a jumbo phage, so Anthropic's claim is narrower: Claude "appears to be the first to notice" the repeat array and a second protein that sits beside the enzyme.

Why CRISPR comes up

A CRISPR array holds a bank of different RNA sequences, and that bank is what makes CRISPR tools programmable. Anthropic's first experiments show the ART array is also expressed as a set of distinct short RNAs. That's a hint, and Anthropic stops there. I was surprised how plainly it says, on its own site and on X, that it doesn't know what the system does.

Anthropic@AnthropicAI · Anthropic

Claude has discovered a previously unknown enzyme system hidden in the DNA of bacteriophages. Beside the enzyme’s gene sits a long array of repeating DNA—a structure that looks somewhat similar to CRISPR. We don’t yet understand what this system does, but only a handful of known systems share its features, and all of them are able to cut, copy, and paste DNA. Historically, the discovery of such programmable systems has helped revolutionize medicine. CRISPR, for instance, is now the foundation of genetic medicines. But it will take much more work to learn what this system does, and whether it can be put to similar use. Read more: https://t.co/RuEosScSMb

View the post on X

Feng Zhang, a CRISPR pioneer who teaches at MIT and the Broad Institute, reviewed the preprint. Anthropic printed his reaction on its page, and it's measured.

The identification of RNA-repeat arrays associated with reverse transcriptases is genuinely intriguing and merits further investigation.

From Claude discovers a novel enzyme system \ Anthropic

The lab behind it

The group formed in the spring of 2026. Its lab is in the Bay Area, works only at biosafety levels BSL-1 and BSL-2, and doesn't handle pathogens that can infect humans. Human scientists do all the bench work. (Anthropic says it has tried AI for lab work too, but that suited the ad hoc experiments here less well.)

The agents ran in Claude Science and Claude Code, the same tools any scientist can use, and sometimes in an in-house harness that runs many Claude sessions at once. Anthropic has posted a technical report and says it wants proposals from scientists who'd like to try the approach on their own questions.

More on Anthropic

All Anthropic stories