Claude Discovers Novel CRISPR-like Enzyme System
Anthropic announced that Claude discovered a new enzyme system named ART in phage DNA, featuring a reverse transcriptase, an unknown partner gene, and a repeat array similar to CRISPR. The discovery was made by 950 Claude agents analyzing over 200,000 reverse transcriptases in 21 hours, and was experimentally validated by Anthropic's wet lab.
Key insights
Discovery Process by Claude Agents
Anthropic's life sciences team gave Claude a broad prompt to find interesting new reverse transcriptases in DNA databases. Approximately 950 Claude agents ran for 21 hours, consuming 210 million tokens. They filtered over 200,000 reverse transcriptases down to 3,500 candidates, then selected 20 for human-readable reports. One agent recognized the repeat array in raw sequence, counted repeats, measured spacing, compared with known systems, and confirmed it was novel before submitting.
ART System Structure and Similarity to CRISPR
The system, named ART (Array-associated Reverse Transcriptase), consists of a reverse transcriptase, an unknown partner gene, and a long array of evenly spaced repeated DNA sequences. This array is transcribed into distinct short RNAs, similar to CRISPR's guide RNA array. CRISPR pioneer Feng Zhang noted that such arrays warrant further study. The precise function of ART is not yet known.
Potential as a Gene Editing Tool
Systems with similar features (reverse transcriptase, repeat array, partner gene) are extremely rare and all previously known ones can cut, copy, or paste DNA. CRISPR has led to gene editing drugs, and others are being developed into tools. Anthropic acknowledges ART's function is unclear but suggests it could represent a new gene editing mechanism. Dario Amodei compared the discovery to work he would be proud of as a PhD student.
Anthropic's Life Sciences Team and Wet Lab
Anthropic formed a life sciences research group in spring 2026, with a BSL-1/BSL-2 wet lab in the Bay Area. All experiments were conducted by human scientists. The team uses publicly available Claude Science and Claude Code, meaning other labs can replicate the AI-driven hypothesis generation and experimental validation pipeline. Anthropic is soliciting collaboration proposals from external scientists.
Dario Amodei's Vision for AI in Biology
Amodei stated that AI performance in intellectual domains follows an exponential trend from weak to superhuman over a few years, citing math progress from 2023 to 2026. He believes biology is on a similar trajectory. He argued that AI's utility in biology is not limited by the need for experimentation, as humans can collaborate with AI to perform experiments. He reiterated the goal of curing most diseases in 5-10 years if AI is applied to every stage of the biomedical pipeline.
Comparison to Independent Stanford Discovery
A Stanford team independently described a novel RT system with a non-coding array that is in some ways similar to ART, though the systems evolved independently. Amodei believes we are at the very beginning of finding such systems and developing them into powerful biotech tools.