Quick Take
  • Anthropic says Claude has discovered a previously unknown enzyme system in the DNA of viruses that infect bacteria.
  • Earlier studies had recorded the enzyme at its core but appear to have missed what surrounds it.
  • The findings stand as key evidence that AI can now spot patterns human experts overlooked.
  • Many discoveries that changed medicine began with a scientist noticing something odd in nature.

What Happened

Many discoveries that changed medicine began with a scientist noticing something odd in nature. Scientists’ use of restriction enzymes, found in bacteria, helped launch the biotechnology industry.

The agent was one of roughly 950 Claude agents working on the same search. Anthropic’s scientists launched it with one prompt, asking Claude to find new reverse transcriptases in a massive DNA database.

Market Context

Anthropic says Claude has discovered a previously unknown enzyme system in the DNA of viruses that infect bacteria. Earlier studies had recorded the enzyme at its core but appear to have missed what surrounds it.

Anthropic published the findings on September 23. The findings stand as key evidence that AI can now spot patterns human experts overlooked. But what exactly does this discovery mean?

Why It Matters

If he is right, Amodei says faster discovery could reveal new drug targets and new kinds of treatments. It would not shorten clinical trials, but it could send far more promising candidates into the drug pipeline.

He has previously written that AI could help cure most diseases in 5 to 10 years. He calls that goal just barely possible.

Details

Why Should Anyone Outside a Lab Care?

An enzyme from a Yellowstone hot spring became the basis for PCR. That DNA-copying method now sits behind much of modern medical testing.

Claude has now taken that noticing step with only broad direction from Anthropic’s scientists. Anthropic CEO Dario Amodei argues this fits a wider pattern. In 2023, AI models struggled with high-school math.

By late 2026, he says, they are beginning to solve some of the hardest open problems in the field. He believes AI for biology is on a similar curve.

For now, human scientists still run every experiment, and the new system’s function remains unknown.

What Exactly Is CRISPR?

Bacteria face constant attacks from viruses called bacteriophages, or phages. Many bacteria defend themselves with a system called CRISPR.

CRISPR works like a memory bank. Bacteria store short snippets of DNA from past invaders between repeating sequences, and scientists call that stretch an array.

Each snippet is copied into a short piece of RNA. That RNA then guides a protein to find and cut matching viral DNA if the same virus returns.

Scientists noticed CRISPR as an unusual repeat pattern in bacterial DNA. They later learned to swap in RNA guides of their own design, which made the system programmable and turned it into a gene-editing tool now used in medicine.

What Did Claude Find?

The system Claude found is built around a different kind of enzyme called a reverse transcriptase. It copies RNA back into DNA, and bacteria use many such enzymes to fight off viruses, according to Anthropic.

Researchers had already recorded this particular enzyme in a jumbo phage, an unusually large virus that infects bacteria. Earlier studies, however, appear to have overlooked the system’s defining features. Claude seems to be the first to spot them.

One is a partner protein whose job nobody knows yet. The other is a long array of evenly spaced DNA repeats, laid out much like a CRISPR array. CRISPR mostly sits in bacteria, while ART turns up mainly in the phages that infect them.

Anthropic calls the three-part setup array-associated reverse transcriptases, or ART. The Claude agent that spotted it logged its surprise as it read the raw DNA.

Follow us on X to get the latest news as it happens

How Did The Claude Agent Find It?

Over 21 hours, the agents used 210 million tokens, the small chunks of text that AI models process. They gathered more than 200,000 reverse transcriptases and picked out 3,500 new candidate systems.