Anthropic says Claude agents spotted a new CRISPR-like enzyme system — but what it does is still unknown

Anthropic announced on 23 September that it has set up a life-sciences research group with its own laboratory, and shared its first result: its Claude models, working largely on their own, flagged a previously undescribed enzyme system in the DNA of bacteriophages — viruses that infect bacteria.

What was found

The company calls the system “array-associated reverse transcriptases”, or ART. It has three parts: a reverse transcriptase (an enzyme that copies RNA into DNA), a partner gene next to it, and a long array of evenly spaced DNA repeats. That layout resembles a CRISPR array — the feature that makes CRISPR-Cas systems programmable. The reverse transcriptase itself had been described in earlier studies of a jumbo phage; according to Anthropic, Claude appears to be the first to notice the associated repeat array and the extra protein.

How it was found

Per the company, about 950 Claude agents worked for 21 hours and used 210 million tokens. They gathered more than 200,000 reverse transcriptases, picked out 3,500 candidate systems and narrowed them to 20 for detailed reports. Human scientists wrote the initial prompt and did all of the laboratory work.

What is not known yet

Anthropic stresses that ART’s function is unknown. Its first experiments show that the array is expressed as a set of distinct short RNAs, but, as The Next Web notes from the preprint, the team has not yet shown that the enzyme is active or that it acts on those RNAs. There is no evidence so far of a new gene-editing tool, and the preprint has not been peer reviewed.

CRISPR pioneer Feng Zhang (MIT and the Broad Institute), who reviewed the preprint, called the finding “genuinely intriguing” and said it merits further investigation.

Why it matters

Independent replication and functional tests will decide whether ART becomes a useful discovery. For now the main news is the method: AI agents screening huge sequence databases and handing candidates to human scientists to test.