Anthropic reports that its Claude model identified a novel enzyme system within DNA databases, a finding the company describes as a breakthrough from its new Bay Area biology lab. Approximately 950 AI agents spent 21 hours scanning over 200,000 reverse transcriptases, enzymes responsible for converting RNA into DNA. This process revealed repeating sequences next to an unusual enzyme dubbed ART, which appears primarily in bacteriophages, viruses that infect bacteria. While human researchers set the initial question and conducted subsequent laboratory tests, the computational analysis remained largely autonomous. The specific function of this system is currently unknown.
The scientific community reaction highlights a tension between marketing narratives and established research methods. Lucas Harrington, who co-founded Mammoth Biosciences with Jennifer Doudna, argues that the technique represents routine genome mining rather than a genuine discovery. Harrington notes that similar systems have been known since 2008 and that the difficult work involves determining biological function, which Anthropic has not yet demonstrated. Presenting early computational results as a major scientific milestone risks inflating the perceived significance of the work.
- The system consists of repeating DNA sequences found near an unusual enzyme.
- Function of the ART system remains unverified by wet-lab experiments.
- Similar genetic systems were identified in public databases as early as 2008.
