SAN FRANCISCO, September 25 — Anthropic says one of its AI models has done something that, until now, was the work of human researchers: it found a new enzyme system in a vast database of DNA, one that looks a lot like the raw material behind CRISPR. The company published a research post on September 23 describing how its model, Claude, autonomously spotted the system.
The finding came out of a research program run by a new life sciences group and laboratory at Anthropic, formed in the spring of 2026 to test a straightforward question: can a general AI model systematize and accelerate biological discovery?
What Claude was asked to do
The team gave Claude a prompt to search a massive database of DNA sequences for interesting new examples of reverse transcriptases, or RTs. Think of RTs as molecular scribes — enzymes that copy RNA back into DNA. That’s the opposite of the usual direction genetic information flows, and it’s a trick that has made RTs useful in labs for decades.
From there, the humans stepped back. According to the post, human involvement was limited to the initial prompt and the lab work.
Roughly 950 Claude agents searched the data for 21 hours, burning through 210 million tokens. One agent noticed something odd: a repeating pattern of DNA sequences sitting next to the gene for a strange-looking RT.
From a hunch to a lab result
Further analysis and testing in Anthropic’s lab confirmed the pattern marked a previously uncharacterized enzyme system found in bacteriophages — viruses that infect bacteria. The team calls it array-associated reverse transcriptases, or ART. The underlying RT, found in a jumbo phage, had turned up 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 sequences and an additional accessory protein of unknown function. Here’s the part that has biologists paying attention.
The ART system carries a set of characteristics that have only ever been found together in a handful of other systems, all of which are programmable and perform operations such as cutting, copying, and pasting DNA. CRISPR is the famous member of that club. Several others are in development as promising tools.
Anthropic released a pre-print with more detail on the work. Feng Zhang, a CRISPR genome editing pioneer and a professor at MIT and the Broad Institute, reviewed it.
Why a software company is doing biology
Anthropic’s life sciences organization includes teams working on drug discovery and on training Claude in biology and chemistry. The lab sits in the Bay Area and conducts research only at biosafety risk levels BSL-1 and BSL-2. It does not handle pathogens that can infect humans, and all lab work is performed by human scientists.
The company frames the discovery against a long history of biology breakthroughs that started with odd observations in nature. Restriction enzymes, found in bacterial immune systems where they destroy invading viral DNA, launched the biotechnology industry.
Taq polymerase, identified in a bacterium in a Yellowstone hot spring, became the basis for PCR. CRISPR itself was first noticed as an unusual repeat sequence in the DNA of certain bacteria and is now the foundation of gene editing-based medicines. Anthropic says its team’s research before joining the company helped better understand the evolution and regulation of CRISPR systems, discover new enzymes for next-generation cell and gene therapies, and build tools for accelerating the identification of anomalies in DNA, such as human pathogenic variants.
The company said its work to understand the primary function of ARTs is ongoing, and that it shared the findings early both to demonstrate Claude’s capabilities and to give the broader community insight into its work. What to watch next: whether Anthropic’s lab pins down what ART actually does, and whether the approach — point a model at a database, let it hunt for anomalies, then verify at the bench — becomes a routine first step in enzyme discovery.
