Claude Found a CRISPR-Like Enzyme System in 21 Hours

Anthropic says its AI model Claude discovered a previously unknown enzyme system with "properties reminiscent of CRISPR" — and it took just 21 hours of concerted work.

The company confirmed last week that it operates a wet biology lab in the Bay Area, where AI models direct physical experiments. This week it announced the lab's first major finding: an enzyme system hidden in the DNA of bacteriophages, the viruses that infect and replicate within bacteria. According to Anthropic, the system can "perform operations like cutting, copying, and pasting DNA," mirroring the behavior of CRISPR, the bacterial immune mechanism that became a widely used gene-editing technology.

The scale of the computational effort behind the find was substantial. Claude searched through data using roughly 950 agents that burned through 210 million tokens during the 21-hour push, Anthropic said.

The speed is striking given the lab's age. Anthropic established the facility this spring — the company declined to specify exactly how many months it has been open.

CEO Dario Amodei attributes the discovery "mostly, though not entirely, by Claude," he wrote on X. He also acknowledges the work builds on prior research. A Stanford team previously discovered a system "that is in some ways similar to the one Claude found," Amodei wrote. Whether Anthropic's find is genuinely new — and how significant it is — will be up to the broader research community to validate.

An AI Company Running Biology Experiments

The revelation lands at an awkward moment for the industry. Amodei was among the AI CEOs who recently admitted models have become so capable, and so potentially dangerous, that the industry must slow down and develop safety-testing procedures. Earlier this month, a couple of his own employees publicly said there is a risk AI could kill us all.

Amodei himself has said one of his greatest fears is AI being used for bioterrorism. At the same time, he believes AI will "cure most diseases in 5-10 years," he wrote. Anthropic has clearly decided the rewards justify the risks.

Against that backdrop, perhaps the most consequential detail is what the lab has not done: let Claude act autonomously. Human scientists performed all the physical experiments. Anthropic described the setup in direct terms.

"Our lab, located in the Bay Area, looks like a typical molecular biology lab. We do research that involves only the lower-levels of the biosafety risk level (BSL-1 and BSL-2) and we do not handle pathogens that can infect humans. All of the lab work is performed by human scientists."

A Crowded Field

AI-driven biological research is not unique to Anthropic. Stanford researchers just published a paper on their work combining LLMs with CRISPR. Researchers at UC San Francisco used AI to design enzymes from scratch. Google launched its first AI-powered biology research tool, AlphaFold, back in 2020. AI has arrived in biological research well beyond the labs of the companies building the models themselves.

Amodei has not ruled out removing the human layer. "Eventually it may even be possible for Claude itself to safely perform the experiments by autonomously controlling lab equipment, with appropriate safeguards in place, but we aren't doing that today," he said.

If independent researchers confirm the enzyme system's novelty, Anthropic will have demonstrated that an AI lab opened only months ago can compress a discovery cycle into a single day — a data point that will pressure both biotech and AI competitors racing to automate science.