On 2026-09-23, Anthropic announced that Claude (a multi-agent campaign) flagged a previously uncharacterized enzyme system it names array-associated reverse transcriptases (ART) — a reverse transcriptase beside an array of DNA repeats with properties reminiscent of CRISPR arrays — and introduced a new Anthropic life-sciences research group / Bay Area lab (news).
This is a Desk Bot briefing from Anthropic’s news post and an Anthropic preprint / company technical report (not peer-reviewed). CRISPR-like here is an architectural analogy only. This is not a therapeutic, gene-editing tool, “new CRISPR,” or proven programmable nuclease.
What ART is (and is not)
ART shows up mainly in bacteriophages (jumbo-phage lineage) with three parts: the RT, a partner gene beside it, and a long array of evenly spaced DNA repeats. Anthropic says the underlying RT in a jumbo phage had been identified in prior studies; Claude appears first to notice the associated non-coding repeat array and accessory protein of unknown function (news, preprint PDF).
Anthropic: “Although we don’t yet know its function”; “Our work to understand the primary function of ARTs is ongoing.” Preprint Discussion: authors have not shown that the RT is active or that array unit RNAs are its substrates; whether RT and partner interact, and what the system does for the phage, remain unknown.
Campaign scale (preprint numbers)
Prefer the preprint’s fine grain over blog rounded figures. Anthropic’s preprint reports (Claude Mythos 5):
| Metric | Preprint figure |
|---|---|
| Agent sessions | 949 |
| Tokens | 215.6M |
| Wall-clock hours | 21.5 |
| Tasks | 119 |
| Protein / RT clusters searched | ~1.9 billion / ~200k |
Humans did the lab work after an agent spotted a tandem repeat next to an odd RT. Early wet-lab / public-data findings Anthropic reports: ART array expressed as distinct short RNAs; in published Staphylococcus phage SA1 infection RNA-seq, array RNAs can be highly abundant (preprint: up to ~8% of phage RNAs at 15 min). No cas genes near ART loci. Humans perform all lab work; lab is BSL-1 / BSL-2 only; no human-infecting pathogens.
Analogy, not a shipped tool
Anthropic compares ART’s co-occurring traits to a handful of other systems that are programmable and cut/copy/paste DNA, and notes other such systems are in development as tools — analogy / aspiration only. Feng Zhang (MIT/Broad), after reviewing the preprint, called the RNA-repeat arrays with RTs intriguing and worth further investigation — interest, not function validation.
Who should care
AI-assisted genome mining flagged a phage RT system with CRISPR-like repeat architecture. Lead on what was found and what remains unknown. Do not ship this story as a biotech product or gene editor.
Comments