Claude finds novel enzyme in DNA: Could AI be entering the gene-editing discovery race? 

Artificial intelligence is moving deeper into biological research, with Anthropic saying its Claude AI system has independently identified a previously uncharacterised enzyme system that has features reminiscent of CRISPR. 

The company said on September 23 that Claude discovered the system after analysing vast amounts of DNA data and scientific literature. The finding has not yet been shown to have a practical gene-editing function, and researchers are still working to determine what the system actually does.

But Anthropic says the discovery demonstrates how AI could help scientists identify biological mechanisms that may otherwise remain hidden in enormous genomic datasets.

The system, which Anthropic has named array-associated reverse transcriptases (ARTs), is primarily found in bacteriophages — viruses that infect bacteria. It is built around a reverse transcriptase, an enzyme capable of copying RNA into DNA, along with a neighbouring partner gene and a long array of regularly spaced DNA repeats. 

MUST READ | Anthropic says Claude now leads 26% of its AI R&D work: How fast is AI development accelerating?

The repeat structure caught Claude's attention because it resembled the arrays associated with CRISPR systems. Anthropic stressed, however, that ART is not being presented as a new CRISPR system or as an established gene-editing technology. Its biological function remains under investigation.

But there is a crucial caveat: finding a biological system is not the same as proving that it has a useful medical or biotechnology application. 

How Claude found it 

Anthropic's researchers initially asked Claude to search a massive database for unusual examples of reverse transcriptases. Rather than examining the data as a single AI system, roughly 950 Claude agents worked in parallel. 

Over about 21 hours, the agents used approximately 210 million tokens to examine the data. They identified more than 200,000 reverse transcriptases, selected about 3,500 candidate systems for further analysis and eventually narrowed the field to 20 of the most compelling candidates.

IN CASE YOU MISSED IT | AI-driven drug discovery race: Anthropic quietly opens biology laboratory in major push

One agent then detected an unusual pattern in DNA located next to a reverse transcriptase. It identified a tandem repeat array that had not previously been recognised as part of a biological system. 

Claude subsequently counted the repeats, examined their spacing, compared the structure with known reverse-transcriptase systems and searched the scientific literature for previous descriptions. The candidate was then passed to Anthropic's scientists for experimental testing. 

Humans still did the laboratory work 

The discovery is significant not simply because an AI found an unusual DNA pattern, but because the process extended from data mining and hypothesis generation into experimental validation. 

Anthropic says its scientists carried out the laboratory experiments and are continuing to investigate ART's function. Initial experiments indicate that the DNA repeat array is transcribed into a collection of short RNAs — a finding that raises the possibility that the repeats could have a functional role analogous, in some respects, to repeat-associated components of CRISPR systems.

Feng Zhang, a CRISPR pioneer and professor at MIT and the Broad Institute, described the identification of RNA-repeat arrays associated with reverse transcriptases as intriguing and said it merits further investigation.

DON'T MISS | 'Poorly managed AI poses existential risk': Tech leaders warn UN, urge global regulation

Anthropic's laboratory operates at BSL-1 and BSL-2 levels and does not handle pathogens capable of infecting humans. The company said all laboratory work in this research programme is performed by human scientists.

Why CRISPR is an important comparison 

CRISPR itself began as an observation of unusual repeated DNA sequences in bacteria. Scientists eventually discovered that these systems form part of bacterial defence mechanisms and learned how to adapt them into programmable tools capable of modifying genetic material. 

That transformation turned a basic biological observation into one of the most important technologies in modern biotechnology. 

FIND WHY | 150,000 animals used each year: How AI could change chemical safety testing

Other biological systems have followed similar paths. Restriction enzymes, for example, were originally part of bacterial defence mechanisms before researchers adapted them for manipulating DNA. Taq polymerase, discovered in heat-loving bacteria, became a crucial component of PCR, the DNA amplification technique widely used in research and diagnostics.

Anthropic's argument is that AI could dramatically expand the search for similarly useful biological systems by examining genomic data at a scale that would be difficult for individual research teams. 

From AI hypothesis to biological discovery 

Dario Amodei, Anthropic's CEO, said the work was conducted "mostly, though not entirely, by Claude". According to Amodei, researchers initially provided a broad research direction, after which Claude searched literature and genomic datasets, identified an interesting system and proposed experiments. Human scientists then performed those experiments to validate the findings. 

The distinction is important: Claude did not independently conduct a laboratory experiment or create a proven gene-editing technology. It identified a biological pattern, developed hypotheses around it and helped direct the subsequent investigation. 

DO CHECKOUT | Claude AI was used in five cases linked to potential bioweapons research, Anthropic says

Anthropic's workflow involves AI searching protein families, identifying unexplained genomic neighbours and generating reports describing possible functions. Scientists then review those candidates, with most being discarded before laboratory testing. Survivors are experimentally characterised, with Claude also helping interpret the resulting data.

The bigger AI-biology bet 

Anthropic sees the discovery as an early example of a broader shift in scientific research. 

The company is building a life sciences organisation around using Claude for tasks ranging from fundamental biology and genomics to drug discovery. Its recently launched Life Sciences Verification Program is also designed to give vetted life-science professionals access to models with safeguards tailored for biology-related work.

FIND WHY | OpenAI warns of autonomous AI risks, calls for global rules before AI starts improving itself

The potential payoff goes well beyond discovering unusual enzymes. Faster identification of biological mechanisms could eventually help researchers identify new drug targets, develop new therapeutic approaches, improve biological measurement and shorten the cycle between computational hypotheses and laboratory experiments. 

Anthropic has not yet established what ART does, whether it can be programmed, whether it can modify DNA in a useful way, or whether it will become a practical gene-editing tool.