Claude Flagged a Phage Enzyme System; Its Function Is Still Open
Anthropic says Claude mined DNA for reverse transcriptases and flagged ART — a known jumbo-phage enzyme plus CRISPR-like repeats and an accessory protein that Anthropic says no one had previously tied together. About 950 agents, 21 hours, 210 million tokens. Function unknown; humans do the bench work.
Corrected October 14, 2026: The summary now attributes to Anthropic the statement that no one had previously tied the enzyme, repeats and accessory protein together.
The claim is a candidate, not a gene-editing tool. What is concrete is the split of labor: agents did the genome mining, people did the bench, and the function is still open.
Anthropic (September 23): a new life-sciences research group and a Bay Area laboratory. Claude, given a prompt to search a large DNA database for interesting reverse transcriptases (enzymes that copy RNA into DNA), flagged a system Anthropic calls array-associated reverse transcriptases (ART).
This is not the September 17 Life Sciences Verification Program (an access grant). This is a research claim from Anthropic’s own lab.
What was new versus what was already known#
The underlying reverse transcriptase, from a jumbo phage, had been identified in earlier studies. Anthropic says Claude appears to be the first to notice the system’s defining features — an associated array of non-coding DNA repeats and an accessory protein of unknown function. The repeat layout resembles a CRISPR array. First experiments show the array is expressed as a set of distinct short RNAs. The primary function is unknown. Anthropic says the combination of traits has only been seen together in a handful of other systems, all of them programmable DNA tools, and it is not claiming ART is one of those tools yet.
Scale (lab-stated)#
After the initial prompt, agents combed the database. One search ran about 21 hours, roughly 950 agents, 210 million tokens. Across the program, agents gathered over 200,000 reverse transcriptases, picked out 3,500 candidate systems, and narrowed those to 20 human-readable reports. Anthropic says its scientists’ role was the initial prompt and the lab work.
Lab limits: BSL-1 and BSL-2 only. No human pathogens. All bench work is done by human scientists. The group says it has experimented with the Model Hardware Standard for automating lab work, and that this style of molecular biology is a poor fit for that.
Outside comment in the post: Feng Zhang (MIT and the Broad Institute), after reviewing the preprint: “This is an exciting example of how AI agents can contribute to biological discovery. The identification of RNA-repeat arrays associated with reverse transcriptases is genuinely intriguing and merits further investigation.” Zhang’s line is an invitation to check the preprint, not a confirmation that ART works like CRISPR.
Preprint: anthropic.com PDF.