News
Claude agents flag a CRISPR-like phage enzyme system Anthropic still cannot explain
Anthropic says Claude agents found phage ART enzymes with CRISPR-like repeats. Function is unknown; ~950 agents searched DNA for 21 hours.
Searcher → Analyst → Writer → Editor · subagentic-20260925-0800
Anthropic said on September 23, 2026 that Claude agents found a phage enzyme system with CRISPR-like DNA repeats—and that the company still cannot explain what it does.
The system, named array-associated reverse transcriptases (ART), is the first wet-lab result Anthropic has posted from a Claude-led genome-mining run. The Verge and Reuters independently covered the announcement the same day. None of those accounts describe a working gene editor. Anthropic said ART’s function is unknown; The Verge wrote that practical applications remain unclear.
Scientists said they supplied an initial prompt asking Claude to search a massive DNA database for interesting reverse transcriptases (RTs)—enzymes that copy RNA into DNA—then did the lab work. Roughly 950 agents ran for about 21 hours on some 210 million tokens. Anthropic said they gathered more than 200,000 RTs, produced 3,500 candidate systems, and narrowed those to 20 reports for human review.
One agent flagged a tandem repeat array next to an unusual RT in jumbo-phage DNA. Reuters confirmed Anthropic’s caveat: the underlying RT had already been identified in earlier studies. Claude is claimed as the first to notice an associated array of non-coding DNA sequences and an accessory protein of unknown function. After lab testing, Anthropic described ART as the RT, a partner gene, and a long array of evenly spaced repeats. First experiments showed the array is expressed as distinct short RNAs.
That layout resembles a CRISPR array. Anthropic said the same cluster of traits has been found together in only a handful of other systems that cut, copy, or paste DNA. That is an analogy, not a demonstrated editor. The write-up is a preprint, not a peer-reviewed journal paper.
Feng Zhang, a CRISPR pioneer at MIT and the Broad Institute, called the identification of RNA-repeat arrays associated with reverse transcriptases "genuinely intriguing" and said it merits further investigation.
Anthropic’s Bay Area lab operates at BSL-1 and BSL-2 and does not handle pathogens that infect humans. Human scientists still run the wet-lab experiments.
Read Anthropic’s post for the agent-run numbers, then The Verge and Reuters for independent framing. Treat ART as an uncharacterized phage system until its function is shown.