Anthropic says its biology lab found a CRISPR-like enzyme system

Anthropic says its biology lab has already produced a discovery it considers significant: a newly identified enzyme 'system' with properties the company describes as 'reminiscent of CRISPR.' The announcement, reported by TechCrunch on September 23, comes just a week after Anthropic confirmed it operates a wet biology lab in the Bay Area where it uses its own AI models to run physical experiments.

According to Anthropic, the finding is a previously unknown enzyme system hidden in the DNA of bacteriophages, the viruses that infect and replicate inside bacteria. The company's researchers say it behaves similarly to CRISPR in that it can 'perform operations like cutting, copying, and pasting DNA.' CRISPR began as a natural bacterial immune system for fighting viruses and became a widely used gene-editing technology.

What has drawn the most attention is how the system was found. CEO Dario Amodei wrote on X that it was discovered 'mostly, though not entirely, by Claude.' By the company's account, Claude needed only 21 hours of concerted effort, searching through data with about 950 agents that burned through 210 million tokens. The lab itself was only established this spring.

Amodei also acknowledged that the work builds on others: a Stanford team previously discovered a system 'that is in some ways similar to the one Claude found,' he wrote. Anthropic concedes it will be up to the broader research community to validate how big, or how new, the discovery actually is.

TechCrunch notes that the biggest reveal may not be the discovery at all, but that the lab has not let Claude run loose. The physical experiments were done by humans. Anthropic describes the lab as looking like a typical molecular biology lab, restricted to the lower biosafety levels BSL-1 and BSL-2 and handling no pathogens that can infect humans: 'All of the lab work is performed by human scientists.'

The timing is delicate. The news lands right after several AI CEOs, Amodei included, publicly argued that models have become capable, and potentially dangerous, enough that the industry must slow down and build safety-testing procedures, and after some Anthropic employees said this month that AI carries a risk of killing us all. Amodei has said one of his greatest fears is AI-enabled bioterrorism, while also saying he believes AI will cure most diseases in five to ten years.

Anthropic is not alone in pushing AI into biology. Researchers at Stanford just published work on LLMs and CRISPR, a UCSF team used AI to design enzymes from scratch, and Google shipped AlphaFold back in 2020. Amodei has not ruled out a fully automated lab: it may eventually be possible, he said, for Claude itself to safely perform experiments by autonomously controlling lab equipment with appropriate safeguards, 'but we aren't doing that today.'