Anthropic said on Wednesday that its Claude AI models autonomously discovered a previously uncharacterized enzyme system hidden in the DNA of bacterial viruses, one of the most concrete examples to date of an AI system driving an original scientific discovery rather than simply accelerating existing workflows. The company announced the finding alongside the launch of a dedicated life sciences research group and laboratory, and published a pre-print paper describing the new system, which Anthropic calls array-associated reverse transcriptases, or ART.
The announcement lands at a moment when frontier labs are racing to prove that AI can generate new knowledge, not just synthesize what humans already know. According to Anthropic, the discovery emerged from a research program formed in the spring of 2026 to test whether general-purpose AI models can systematize biological exploration — sweeping through datasets that no human team could read in a lifetime. For more context on this story, see our ongoing latest AI developments.
How the discovery unfolded
Anthropic said its scientists gave Claude a single high-level prompt: search a massive database of DNA sequences for interesting new examples of reverse transcriptases, enzymes that copy RNA into DNA. From that point on, the company said, the model's agents worked largely on their own.
Roughly 950 Claude agents spent 21 hours combing through the database, consuming around 210 million tokens, according to Anthropic's announcement. One of them flagged something unusual: a repeating pattern of DNA sequences sitting next to the gene for an odd-looking reverse transcriptase carried by a jumbo phage, a giant virus that infects bacteria.
"While this underlying RT had been identified in previous studies, Claude appears to be the first to notice the system's defining features — an associated array of non-coding DNA sequences and an additional accessory protein of unknown function," Anthropic wrote in its announcement. After further analysis and laboratory testing by the company's scientists, the team concluded the pattern marked an enzyme system that had never been described before.
Why the finding matters
The system's characteristics have only ever been found together in a handful of other biological systems, all of which are programmable and capable of operations like cutting, copying, and pasting DNA, according to Anthropic. That family includes CRISPR, the gene-editing tool that has already produced approved medicines, as well as several newer systems now in development as research tools.
Anthropic was careful to note that the function of ART remains unknown. The company said its work to understand what the system actually does in bacteriophages is ongoing, and that it chose to publish early both to demonstrate Claude's capabilities and to give the research community a head start on investigating the finding.
An old story told at new speed
The history of molecular biology is full of transformative tools that began as oddities noticed in obscure organisms. Restriction enzymes, which cut DNA at specific short sequences, were first found in bacterial immune systems and went on to launch the entire biotechnology industry. Taq polymerase, an enzyme that copies DNA at high temperatures, was discovered in a bacterium living in a Yellowstone hot spring and became the foundation of PCR, the amplification method behind much of modern diagnostics. CRISPR itself was first spotted as strange repeating sequences in bacterial DNA long before anyone understood it as an editable immune system.
Anthropic is explicitly framing ART as the latest entry in that lineage — except this time, the company argues, the noticing was done by an AI agent working at database scale. Whether the comparison holds depends on what further research reveals about the system's function, a point scientists will now probe using the pre-print Anthropic released.
The lab behind the discovery
The team behind the work is a new life sciences research group at Anthropic, staffed by scientists who specialize in computational approaches to reading and interpreting DNA. The group operates a molecular biology laboratory in the San Francisco Bay Area where experiments are conducted only at the lowest biosafety levels, BSL-1 and BSL-2, and no pathogens capable of infecting humans are handled, according to the company. Anthropic said all wet-lab work is performed by human scientists, with Claude agents handling the computational exploration, hypothesis generation, and analysis.
That division of labor reflects a broader debate inside the AI industry about how quickly autonomous systems should be inserted into biological research. Anthropic's approach keeps physical experimentation under direct human control while letting agents do what they are best at: exhaustively searching data for patterns no human would think to look for.
Expert reaction
The pre-print drew early praise from Feng Zhang, one of the pioneers of CRISPR genome editing and a professor at MIT and the Broad Institute.
"This is an exciting example of how AI agents can contribute to biological discovery," Zhang said in a statement quoted by Anthropic. "The identification of RNA-repeat arrays associated with reverse transcriptases is genuinely intriguing and merits further investigation. I hope this work encourages more scientists to explore how AI can support their research."
What happens next
Anthropic says its work to understand the primary function of ART is ongoing, and the company is positioning the new research group to keep pairing Claude-driven database exploration with hands-on laboratory testing. If ART turns out to be programmable the way its structural cousins are, it could eventually join the growing toolbox of enzymes used to manipulate DNA in research and medicine — a result that would be remarkable either way, whether it was found by an agent or a postdoc.
For now, the discovery stands as a proof of concept with a simple implication: in fields drowning in data, the bottleneck may no longer be noticing the anomaly. It may be deciding what to do with it.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
