Michael JohnsonAnthropic reported that Claude helped identify an unusual combination of an enzyme and repeating DNA sequences in the DNA of viruses that infect bacteria. The company was intrigued by the resemblance between these repeats and one component of CRISPR. But scientists still don’t know what the system does, and have not shown that it can edit genes, as Anthropic’s account of the discovery explains.
What exactly was found
The system was named ART, short for array-associated reverse transcriptases. Researchers identified three components: the enzyme reverse transcriptase, a neighboring partner gene, and a series of repeating DNA sequences.
Reverse transcriptase is an enzyme that can use RNA as a template to make DNA. The phages whose genomes contain ART are viruses that infect bacteria, not people. According to Anthropic’s description of the system, the enzyme itself was already known; the new observation was its connection to a repeating DNA sequence and a neighboring gene whose function is also unknown.
Why CRISPR came up
ART’s sequences consist of short repeats arranged one after another. This resembles the structure of CRISPR, a system that scientists have turned into a tool for working with DNA. In early laboratory experiments, researchers also found that an ART sequence produces short RNA molecules.
For now, that’s where the resemblance ends. The results do not show that ART recognizes specific sites in DNA, cuts it, or enables targeted changes. Anthropic says the system’s biological role is still unknown and that it is under further study. So calling ART “the new CRISPR” or a ready-made gene-editing tool would be premature, according to the company’s report.
How Claude contributed to the search
Researchers asked Claude to search large databases of biological sequences for unusual reverse transcriptases. According to Anthropic, around 950 Claude agents analyzed data for about 21 hours. One of them noticed a repeating sequence next to an enzyme gene and flagged the finding for people to check.
Researchers then examined the lead and conducted laboratory experiments. In other words, Claude helped spot a possible pattern, but its biological significance has not yet been established. That distinction matters: finding an unusual combination of elements in a genome is not the same as understanding how it works, as Anthropic’s account of the research notes.
What we can conclude for now
For now, ART is an intriguing research finding, not a proven biotechnology. It shows how AI can help sift through large datasets and draw attention to patterns worth investigating. But further conclusions depend on experiments: scientists need to determine what role the repeats and neighboring gene play, whether they interact with reverse transcriptase, and what the system does inside the phage.
Anthropic announced the results on September 23, 2026. The company also published a preliminary description of the research; independent review and further experiments are especially important because ART’s main function remains an open question, as the company’s announcement makes clear.