Last week, Anthropic verified that it runs a wet biology laboratory in the Bay Area, where its AI models are used to conduct physical experiments. Now, the company says that lab has produced what it considers a major breakthrough: a new enzyme "system" with certain "properties reminiscent of CRISPR," as Anthropic puts it. CRISPR is a naturally occurring immune mechanism that bacteria use against viruses, and it has become a widely adopted gene-editing tool among researchers. In essence, Anthropic claims to have uncovered a previously unknown enzyme system concealed within the DNA of bacteriophages—viruses that infect and replicate inside bacteria. According to Anthropic's researchers, this system functions much like CRISPR in that it can "perform operations like cutting, copying, and pasting DNA."

Whether this discovery is truly as significant, or as novel, as it sounds will need to be determined by the wider scientific community. Anthropic CEO Dario Amodei concedes that the finding built on earlier work by others, and he noted on X that a Stanford team had previously identified a system "that is in some ways similar to the one Claude found." But Amodei and Anthropic aren't focused solely on the discovery itself. They're also highlighting that it was made "mostly, though not entirely, by Claude," as the CEO put it. That is remarkable on its face. The lab was only set up this spring (Anthropic declined to specify exactly how many months it has been running). A discovery within just a few months would already be fast work, but the team says Claude actually needed only 21 hours of focused effort. Claude combed through data using roughly 950 agents that consumed 210 million tokens. Yet the news that Anthropic has been running biology experiments lands right after AI CEOs, Amodei included, publicly conceded that models have grown so capable—and so potentially hazardous—that the industry needs to slow down and build safety-testing protocols. This follows a couple of his own employees saying publicly earlier this month that AI could pose an existential risk to humanity. Amodei himself has said one of his greatest fears is that AI could be weaponized for bioterrorism. But he also believes AI will "cure most diseases in 5-10 years," he wrote. So Anthropic has evidently concluded that the potential benefits outweigh the dangers. With that in mind, maybe the most telling detail in this news isn't the discovery at all, but the fact that the biology lab hasn't given Claude free rein. The physical experiments were carried out by people. "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."

It's also worth noting that AI-driven biological research is far from exclusive to Anthropic. Stanford researchers just published a paper on their work combining LLMs and CRISPR. Scientists at UC San Francisco used AI to design enzymes from scratch. Google introduced its first AI-powered biology research tool, AlphaFold, back in 2020. So AI has firmly established itself in biological research well beyond the AI labs developing it. Even so, Amodei hasn't dismissed the idea of a fully automated biology lab down the road. "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.