Anthropic’s Claude AI Discovers CRISPR-Like Enzyme System in Bacteriophage DNA

Anthropic says its Claude AI model has identified a previously undescribed enzyme system hidden in the DNA of bacteriophages, highlighting the growing role of artificial intelligence in biological research.

The system, which Anthropic has named array-associated reverse transcriptases (ART), was discovered after Claude autonomously searched a database containing approximately 1.9 billion protein clusters.

Researchers say ART includes a reverse transcriptase enzyme, a neighboring partner gene and a long sequence of evenly spaced DNA repeats. The arrangement has similarities to the repeat arrays found in CRISPR systems, although Anthropic has not yet determined what the newly identified system actually does.

The discovery was reported by Anthropic alongside a preprint detailing the AI-assisted research process.

Claude AI Searches 1.9 Billion Protein Clusters

The discovery came from an autonomous research campaign in which Claude analyzed a massive collection of protein information.

According to the preprint, the campaign ran for approximately 21.5 hours and involved 949 AI agent sessions.

During the process, the agents used approximately 215.6 million tokens and analyzed large numbers of enzyme and protein clusters.

The AI system initially identified around 200,000 enzyme clusters. Researchers then narrowed the search to approximately 3,564 candidate partner families, eventually producing 19 reports for human researchers to examine.

The ART system emerged during this investigation when one AI agent examining DNA near an unusual enzyme noticed a distinctive repeat pattern in the surrounding sequence.

The observation ultimately led researchers to examine the combination of the enzyme, neighboring gene and repeated DNA array more closely.

What Is the ART Enzyme System?

Anthropic refers to the newly identified system as array-associated reverse transcriptases, or ART.

The system consists of three notable components:

• A reverse transcriptase enzyme
• A partner gene located next to the enzyme
• A long array of regularly spaced DNA repeats

The repeat arrangement is particularly interesting because it resembles the structure of a CRISPR array.

CRISPR systems use arrays containing genetic sequences that can serve as guides for targeting specific DNA or RNA sequences. Their programmable nature has made CRISPR an important technology in modern genetic research.

However, the similarity between ART and CRISPR does not mean that ART performs the same function.

Anthropic has emphasized that researchers do not yet know what the ART system does.

Claude Identified More Than Researchers Had Previously Noticed

The reverse transcriptase component associated with ART was not completely unknown.

Anthropic said the enzyme itself had been identified in previous research. What Claude apparently brought to the researchers’ attention was the combination of the enzyme with the previously overlooked repeat array and an adjacent partner protein whose function remains unknown.

That distinction is important.

The discovery is not necessarily the identification of an entirely new enzyme. Instead, the AI system helped identify a previously undescribed genetic system and its surrounding organization.

This illustrates one potential advantage of AI-assisted biological research: machines can search enormous datasets for patterns that might be difficult for researchers to identify manually.

CRISPR Pioneer Feng Zhang Reviews the Finding

The findings have attracted attention from researchers familiar with CRISPR and molecular biology.

Feng Zhang, a professor at MIT and the Broad Institute and a prominent CRISPR researcher, reviewed the preprint.

Zhang described the identification of RNA-repeat arrays associated with reverse transcriptases as intriguing and said the finding warrants additional investigation.

His comments add scientific interest to the discovery while also reflecting the preliminary nature of the research.

Further experiments will be required to establish the biological role of ART and determine whether the system has any useful applications.

Researchers Caution Against Calling ART a New CRISPR

Not everyone views the discovery as evidence of a potential CRISPR replacement.

Kevin Blake, a microbiologist at Washington University School of Medicine, questioned whether the system should currently be viewed as a rival to CRISPR technology.

There is also no evidence at this stage that ART could be developed into a therapeutic tool.

A major reason is that researchers have not yet determined the biological function of the system.

Nature contains an enormous number of unexplored genetic sequences, particularly because many bacterial species have not been extensively studied. As more microbial genomes are analyzed, researchers are likely to uncover many previously unknown genetic systems.

That means the existence of a CRISPR-like genetic arrangement does not automatically indicate that it can be transformed into a useful biotechnology.

Claude’s Discovery Was Not Reproduced in Follow-Up Searches

One of the most important limitations of the finding is its reproducibility.

Anthropic said it repeated the search 10 additional times, but none of those subsequent attempts identified the same repeat array.

The company attributed the outcome to the enormous scope of the search and the variable way autonomous AI agents navigate large research tasks.

The inability to reproduce the result through the same automated search process does not necessarily invalidate the original observation, but it does highlight the need for independent analysis and experimental confirmation.

Human researchers will ultimately need to verify the genetic system and determine whether the observed arrangement has a biological function.

AI Is Taking a Larger Role in Biological Discovery

The ART finding represents an example of how AI systems could potentially assist scientists with large-scale biological research.

Modern databases contain enormous quantities of genetic and protein information. Searching through these datasets manually can be extremely time-consuming.

AI agents can potentially help researchers:

• Search large biological databases
• Identify unusual genetic patterns
• Compare protein families
• Group related sequences
• Generate research hypotheses
• Prioritize candidates for laboratory testing
• Summarize complex biological information

The ART discovery demonstrates one possible workflow in which AI performs large-scale computational exploration while human scientists evaluate and experimentally investigate the resulting candidates.

Anthropic’s Life Sciences Research Expands

Anthropic established its life sciences research group in spring 2026 as part of its broader effort to explore applications of AI in scientific research.

The company’s Bay Area laboratory operates at biosafety levels 1 and 2, according to Anthropic.

The company also said it does not work with pathogens capable of infecting humans.

Importantly, Anthropic said that all laboratory work is conducted by human scientists.

The AI system is being used for computational research and discovery rather than independently conducting physical biological experiments.

What Happens Next?

The biggest unanswered question surrounding ART is its biological function.

Researchers now need to determine whether the repeat array and associated proteins perform a meaningful role in bacteriophage biology.

Further experiments could reveal:

• What the reverse transcriptase actually does
• What role the partner protein plays
• Whether the repeat sequences produce functional RNA
• Whether the system interacts with bacterial or viral DNA
• Whether ART provides any biological advantage
• Whether the system has potential applications in biotechnology

Anthropic said experiments aimed at understanding ART are ongoing.

The company has also invited other researchers to submit proposals for further investigation.

Could ART Become a New Gene-Editing Tool?

It is too early to know whether the newly identified system could eventually become useful in biotechnology.

The comparison with CRISPR is based primarily on the presence of a repeated genetic array and should not be interpreted as evidence that ART can perform CRISPR-like gene editing.

For a newly discovered biological system to become a practical technology, scientists would first need to understand its mechanism and demonstrate that it can be reliably controlled.

If ART turns out to have a programmable or otherwise useful biological function, it could potentially become the subject of future biotechnology research.

For now, however, that remains an open question.

AI and the Future of Scientific Discovery

The discovery highlights a broader shift in scientific research.

AI systems are increasingly being used to analyze datasets that are too large or complex for researchers to examine efficiently by hand. In biology, this could mean searching through billions of protein sequences, predicting molecular structures and identifying unusual genetic relationships.

But AI-generated discoveries still require human verification.

In the case of ART, Claude helped identify a potentially interesting genetic pattern, but researchers must now determine whether that pattern represents a functional biological system and what role it plays.

The combination of AI-powered discovery and human laboratory validation could become an increasingly important model for future scientific research.

Key Takeaways

• Anthropic says Claude AI identified a previously undescribed enzyme system in bacteriophage DNA.
• The system is called array-associated reverse transcriptases (ART).
• ART contains a reverse transcriptase, a neighboring partner gene and a long DNA repeat array.
• The repeat arrangement resembles a CRISPR array, but its function remains unknown.
• Claude searched a database containing approximately 1.9 billion protein clusters.
• The autonomous campaign lasted 21.5 hours across 949 agent sessions.
• The research process used approximately 215.6 million tokens.
• Around 200,000 enzyme clusters were examined before researchers narrowed the candidates.
• Follow-up searches reportedly failed to reproduce the same repeat-array discovery.
• Human scientists are conducting the laboratory work needed to determine ART’s function.
• Researchers caution that there is currently no evidence ART is a CRISPR replacement or a therapeutic technology.

Conclusion

Anthropic’s reported discovery of the ART enzyme system offers an intriguing example of how AI could help scientists search through enormous biological datasets and identify patterns that might otherwise remain unnoticed.

Claude’s role was particularly notable because the model helped connect a known reverse transcriptase with a nearby partner gene and an unusual repeat array resembling a CRISPR structure.

But the discovery is still at an early stage.

Researchers have not yet determined what ART does, and the failure to identify the same repeat array in 10 subsequent autonomous searches highlights the need for further validation.

For now, the most significant takeaway may be less about creating a new CRISPR technology and more about demonstrating how AI agents can assist scientists in exploring the vast and largely unmapped world of biological data.