Prompt and Model
Regulation

Anthropic's AI lab finds new CRISPR-like

Anthropic says its AI model Claude discovered a new enzyme system with CRISPR-like properties in just 21 hours of analysis.

Anthropic says its AI model Claude discovered a new enzyme system with CRISPR-like properties in just 21 hours of analysis

Anthropic has announced its AI-powered biology lab has made a significant discovery. The company says its Claude model identified a previously unknown enzyme system hidden in bacteriophage DNA.

According to Anthropic, this system exhibits properties reminiscent of CRISPR, the gene-editing technology derived from a bacterial immune system. The researchers state the newly found system can perform operations like cutting, copying, and pasting DNA.

The AI-driven discovery

The discovery was made at Anthropic's wet biology lab in the Bay Area, which opened this spring. CEO Dario Amodei said the find was made "mostly, though not entirely, by Claude." The AI model accomplished the core analytical work in just 21 hours.

During that period, Claude searched through data using about 950 agents that processed 210 million tokens. The physical experiments to confirm the AI's findings were subsequently conducted by human scientists in the lab.

Context and validation

Amodei acknowledges the discovery builds on the work of others. He noted a team from Stanford previously discovered a system that is in some ways similar. It will be up to the broader research community to validate how big, or new, this discovery actually is.

The announcement follows recent public statements from AI CEOs, including Amodei, about the need to slow down and develop safety-testing procedures due to model capabilities and potential dangers. Some Anthropic employees have publicly stated they believe there is a risk AI could kill us all.

Amodei has said one of his greatest fears is AI being used for bioterrorism. However, he also believes AI will "cure most diseases in 5-10 years."

The lab's safety approach

Perhaps the biggest reveal is that the biology lab has not let Claude operate autonomously. Anthropic emphasized that all physical lab work is performed by human scientists. The company described its facility as a typical molecular biology lab conducting research at the lower biosafety risk levels, BSL-1 and BSL-2.

We do not handle pathogens that can infect humans, the company stated. This controlled approach contrasts with the rapid AI-driven analysis that led to the discovery.

Amodei has not ruled out a fully automated biology lab in the future. 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.

AI's role in biological research

AI-powered biological research is expanding beyond dedicated AI labs. Researchers from Stanford recently published a paper on their work with large language models and CRISPR. Scientists at UC San Francisco have used AI to design enzymes from scratch.

Google launched its first AI-powered biology research tool, AlphaFold, back in 2020. Anthropic's announcement shows that AI has firmly arrived as a tool for discovery in the life sciences. The company's model identified a potential new biological tool in a matter of hours, a process that traditionally could take years.

Related coverage

More from Regulation