Anthropic Claude Agents Find Crispr-Like Genetic System

Anthropic recently announced that a swarm of its Claude AI agents identified a novel enzyme system with features resembling the Nobel Prize-winning gene-editing tool Crispr. Operating simultaneously, approximately 950 Claude agents scanned massive genomic databases to locate the unique genetic sequences in just 21.5 hours. The newly identified system, which Anthropic calls array-associated reverse transcriptases (ART), was found within jumbo phages, which are large viruses that infect bacteria.

To achieve this, the AI agents were prompted to search for interesting examples of reverse transcriptases, which are proteins that copy RNA into DNA. Claude initially flagged more than 200,000 candidates, narrowed those down to several thousand novel proteins, and ultimately isolated the unusual ART family characterized by repeating DNA sequences. While a 2021 paper by microbiologist Jason Gill had previously identified this specific reverse transcriptase, Claude's contribution was recognizing the surrounding repeat sequences that suggest a Crispr-like mechanism.

For biotechnology practitioners, this deployment highlights a massive shift in workflow efficiency. Mining genomic databases manually for these enzymes typically takes months of human effort, meaning Claude's sub-24-hour turnaround represents a dramatic acceleration. However, experts urge caution regarding the system's actual utility. Stanford University professor Le Cong pointed out that physical wet-lab validation is still pending, noting that "the experiments are still in the queue." It remains unproven whether the ART system can actually cut, copy, and paste DNA, or if it functions more like a retron, which has limited gene-editing capabilities.

Anthropic, which recently established its own wet lab for drug discovery, has published a non-peer-reviewed technical report detailing one physical experiment. While Anthropic CEO Dario Amodei has envisioned a future where Claude autonomously runs laboratory equipment, researchers emphasize that translating basic biological discoveries into viable medical therapies remains a decades-long process. For now, the breakthrough proves that AI is highly adept at pattern recognition in vast datasets, even if human scientists must still do the heavy lifting of physical verification.