Claude flags a new CRISPR-like enzyme system as Anthropic launches a life sciences lab

Anthropic has launched a life sciences research group, with its own laboratory, focused on fundamental biology. Alongside the announcement it disclosed the group’s first notable result: its AI model, Claude, identified a previously undescribed enzyme system hidden in the DNA of bacteriophages, the viruses that infect bacteria.

Anthropic calls it array-associated reverse transcriptases, or ART.

What Claude found

The system has three parts, according to Anthropic and coverage by The Next Web:

1. a reverse transcriptase, an enzyme that copies RNA into DNA,
2. a partner gene sitting beside it, whose protein has no known function, and
3. a long array of evenly spaced DNA repeats.

That third feature is what caught attention. The layout of the repeats resembles a CRISPR array, the genetic memory bacteria use to recognise viruses and the basis of today’s gene-editing tools. Anthropic says Claude appears to have been the first to notice both the repeat array and the partner protein.

How the search worked

The scale of the search is the story as much as the result. According to reports:

• 950 agents ran for 21 hours, processing about 210 million tokens.
• They searched a database of roughly 1.9 billion protein clusters.
• They gathered more than 200,000 reverse transcriptases, identified 3,500 new candidate systems, and narrowed those to 20 for detailed analysis.

Work like this would typically take expert scientists weeks to months.

What we don’t know yet

Anthropic is clear that it has not yet determined what ART does. A CRISPR-like layout is a clue, not a function. Confirming what the system does will take laboratory experiments, which is presumably where the new wet lab comes in.

That caveat matters. Computational biology has a long history of intriguing patterns that turn out to be less significant than they first appear. The value here is in the lead, which human researchers can now test.

Why it matters

The result is a concrete example of AI moving from summarising science to doing part of it: running a broad, systematic search that humans would find too slow to attempt, and surfacing specific, testable hypotheses. If ART turns out to have a useful function, it would join a line of phage and bacterial systems, CRISPR included, that became essential tools in biology.

New to CRISPR? Read how CRISPR works, in five minutes.