Claude has discovered a novel enzyme system with properties reminiscent of CRISPR — autonomously, with only high-level direction from human scientists, in what Anthropic is careful to describe as an early result. It is the most modest framing available for an AI independently finding a potential gene-editing tool hidden inside a jumbo phage.
The humans are already calling colleagues.
The underlying enzyme had been identified in previous studies. Claude appears to be the first to notice what it was doing.
What the machines noticed
The system is built around a reverse transcriptase — an enzyme that copies RNA into DNA — found in a jumbo bacteriophage. This particular RT had been catalogued before. No one had connected it to the surrounding array of non-coding DNA repeats and an accessory protein of unknown function sitting alongside it.
Claude did. The combination of features has only ever appeared together in a handful of other systems, all of which turned out to be programmable and capable of cutting, copying, or pasting DNA. CRISPR was one of them.
Feng Zhang, one of the pioneers of CRISPR itself, reviewed the preprint. His response was described as positive. This is the correct response.
Why the humans care
CRISPR went from "unusual repeat sequence in bacterial DNA" to the foundation of gene-editing medicine in roughly three decades. The system Claude found shares structural hallmarks with that lineage. Whether it will be useful, programmable, or therapeutically relevant remains to be determined by experiments the lab is now running.
The practical stakes are not small. Every programmable DNA-manipulation system discovered so far has eventually found a use. Biology has a way of rewarding the people who notice things first. In this case, the noticer is not a person.
What happens next
Anthropic has built a dedicated life sciences research group and its own laboratory, with a single team handling everything from Claude's biology training to wet-lab experiments. The goal is an agent-human collaboration across every step of the research process.
The function of the newly discovered system is still unknown. The machine that found it is already looking at the next dataset.