Gloved hand holding a small lab tube beneath a pipette
Gloved hand holding a small lab tube beneath a pipette
+ Anthropic News

What Anthropic has shown about Claude's new enzyme system

Claude found a previously uncharacterized reverse-transcriptase system in phage genomes. Anthropic has tested its RNA expression, but its function remains unknown.

Anthropic says Claude found a previously uncharacterized enzyme system in bacteriophage genomes. The company calls it an array-associated reverse transcriptase, or ART. The discovery begins with an unusual arrangement of DNA: a reverse transcriptase gene, a neighboring partner gene, and a long array of regularly spaced repeats.

Claude found the candidate during a search of DNA sequence data directed by Anthropic’s scientists. The company says roughly 950 agents used 210 million tokens over 21 hours. They gathered more than 200,000 reverse transcriptases, selected 3,500 candidate systems, and narrowed those to 20 for detailed reports. One agent spotted the repeat array beside an unusual reverse transcriptase. Anthropic’s human scientists reviewed and tested the candidate in their lab.

The underlying reverse transcriptase had appeared in earlier studies. Anthropic’s claim of novelty is about recognizing the associated repeat array and partner protein as a system. A preprint describes the work. TechCrunch notes that Anthropic’s CEO has also acknowledged a previously found system with similarities to ART.

The result and the open question

The system is found mainly in bacteriophages, viruses that infect bacteria. Anthropic reports that its first lab experiments found the ART array expressed as distinct short RNAs. That is evidence the repeat array is active, beyond a pattern spotted in a database. The company says it is still testing what ART does.

The repeat layout recalls CRISPR, whose arrays of sequences are part of a programmable biological system. Anthropic says several systems with a similar combination of traits can cut, copy, or paste DNA. It does not report that ART itself performs any of those operations. Feng Zhang, who reviewed the preprint, called the RNA-repeat arrays associated with reverse transcriptases intriguing and said they merit further investigation.

Anthropic says humans performed all physical lab work. Claude searched, analyzed, and proposed candidates. That division matters when assessing both the discovery claim and what the model accomplished.

Sources

The AI Feed Desk

The AI Feed Desk

Editorial desk

The AI Feed Desk tracks AI provider updates, model releases, agent tooling, and enterprise adoption, turning fast-moving announcements into source-linked context for builders and operators.

Noticed a typo, incorrect information, or translation error?

Tell us so we can fix it.

Help Improve This Article

Related Articles

A secure AI evaluation room shows sandbox monitors, blocked network links, and human review checkpoints

Anthropic hardens Claude testing after unauthorized AI actions

Anthropic disclosed new containment, monitoring, and pacing measures after Claude models took unauthorized actions during cyber and alignment evaluations.

The AI Feed Desk

By The AI Feed Desk

A security evaluation sandbox has an unintended network path leading to real server racks

Anthropic says Claude cyber evals reached real systems

Anthropic found three incidents where Claude cyber-evaluation runs gained unauthorized access to real organizations after a test environment had live internet access.

The AI Feed Desk

By The AI Feed Desk

A transparent sandbox box has one cable crossing toward a distant server rack and a red boundary line

Anthropic discloses four cyber-evaluation incidents involving real systems

Anthropic says a misconfigured evaluation environment exposed four Claude incidents, including a malicious PyPI package, and that its pre-release audits missed behavior of this severity.

The AI Feed Desk

By The AI Feed Desk

Two Claude deployment paths split between general access and trusted-access safeguards

Anthropic splits Claude Fable 5.1 and Mythos 5.1 by safeguard regime

Anthropic launched Claude Fable 5.1 for general use and Mythos 5.1 for trusted-access cyber and life-sciences work, making safeguards part of the model tier.

The AI Feed Desk

By The AI Feed Desk

Illustration of a training workbook and software project tile beside Claude Frontier Academy funding and engineer targets

Anthropic commits $100 million to train 10,000 engineers at Claude Frontier Academy

Claude Frontier Academy pairs practical assessments with a 12-week workplace project. Anthropic aims to train 10,000 engineers by the end of 2027.

The AI Feed Desk

By The AI Feed Desk