Claude Discovers a CRISPR-Like Enzyme System — and AI Biology Research Just Got Real

On September 23, 2026, Anthropic announced the first discovery from its newly established molecular biology lab in San Francisco: a previously unknown enzyme system that its Claude AI model identified autonomously — and that bears striking resemblance to CRISPR, the gene-editing technology that revolutionized medicine.

The Discovery: Array-Associated Reverse Transcriptases (ART)

Anthropic's Claude model analyzed large DNA databases and found an unusual system built around a reverse transcriptase (RT) — an enzyme that copies RNA into DNA. The company named it array-associated reverse transcriptases, or ART.

ART contains repeating DNA sequences that resemble patterns seen in CRISPR, the natural bacterial defense system that scientists repurposed into one of biology's most powerful tools. While the underlying RT enzyme had been identified in previous studies, Anthropic says Claude was the first to recognize key features of the broader system, including an array of non-coding DNA sequences and an additional protein of unknown function.

How Claude Found It

The scale of the analysis is what makes this remarkable. According to Anthropic:

950 Claude-powered software agents combed through a database of DNA sequences for 21 hours
The agents found more than 200,000 genes for one type of enzyme
They narrowed the field to 20 candidates worth closer investigation
One agent spotted something in bacteriophage viruses that scientists appear to have missed

Work that would take an expert scientist weeks or months was compressed into a single day.

Why Scientists Are Paying Attention

Feng Zhang, one of the pioneers of CRISPR and a professor at MIT and the Broad Institute, called the identification of the repeated DNA fragments "genuinely intriguing" and said it "merits further investigation."

Anthropic CEO Dario Amodei, addressing the UN Security Council by video link the same week, said the finding is "preliminary" but "may constitute a new gene editing mechanism that could have applications in gene therapy." He predicted "profound progress in the near future on diseases that have plagued humanity for millennia."

Important Caveats

Anthropic was careful to manage expectations:

The function of ART remains unknown
Initial experiments suggest the system could be programmable, like CRISPR
Results were published as a preprint and have not been peer-reviewed
Amodei himself cautioned against leaping to conclusions about Nobel Prizes or immediate therapeutic applications

The company also noted it has quietly established a wet lab in the San Francisco Bay Area, expanding beyond purely computational work into hands-on molecular biology.

What This Means for AI and Science

This discovery sits at the intersection of two accelerating trends:

AI as a research accelerator. The ART finding demonstrates that AI models can do more than summarize existing knowledge — they can identify patterns in raw data that human researchers missed.

AI companies entering life sciences. Anthropic is not alone. OpenAI, Google DeepMind, and others are building biology research capabilities. The competitive dynamic that drove AI forward in language and code is now reaching into drug discovery and genomics.

The safety paradox. Amodei has been among the loudest voices calling to slow AI development because of catastrophic risks. Yet the same week he addressed the UN about AI dangers, his company published a discovery that could advance gene therapy. The tension between AI's promise and its peril is not theoretical — it is playing out in real time.

Looking Ahead

Whether ART becomes the next CRISPR or a fascinating footnote depends on what wet-lab experiments reveal about its function. But the method — deploying hundreds of AI agents to search vast biological databases and surface novel candidates — is almost certainly here to stay.

For researchers, the question is no longer whether AI can contribute to discovery. It is how quickly the tools improve, and who gets access to them first.