Anthropic said Claude autonomously discovered a previously unknown enzyme system in the DNA of bacteriophages, with properties the company compares to CRISPR1,4,5. The finding is the first result from Anthropic's wet biology lab, which was established in the Bay Area earlier this year.
Nearly 950 Claude agents worked over 21 hours, processing 210 million tokens, before one agent spotted an unusual repeating pattern in a massive database of DNA sequences. Anthropic said its scientists' involvement was limited to the initial prompt and subsequent lab work to verify the AI's findings. Dario Amodei wrote on X that the work was done "mostly, though not entirely, by Claude": the life sciences team suggested a broad area of research, Claude read through the literature and genome data, discovered something of interest, and then proposed experiments for the team to carry out3.
What Claude found
The enzyme system was hidden in bacteriophage DNA and behaves similarly to CRISPR in that it can perform operations like cutting, copying, and pasting DNA, according to Anthropic. Anthropic described it as a "molecular machine" that it suspects could represent a new gene-editing mechanism. Dario Amodei acknowledged that the discovery built on the work of others, noting that a team from Stanford previously discovered a system that is "in some ways similar".
The company itself hedged on the result's ultimate importance. Dario Amodei said the precise function, biotechnological utility, and level of significance are "not yet clear," but added that "at minimum it is work I would have been proud to do as a PhD student". A Bloomberg headline stated that scientists say the gene-editing discovery "isn't yet a breakthrough"2. TechCrunch noted it will be up to the broader research community to validate how significant or novel the finding actually is.
Amodei's trajectory claim
Dario Amodei used the announcement to argue that AI capability in biology is following an exponential curve analogous to mathematics. He cited a progression: in 2023, models struggled with average high-school-level math; in 2024, they performed well on elite high-school competitions; in 2025, they began solving minor open problems; in early 2026, more significant open problems; and in late 2026, they are beginning to solve "the top few open problems in all of mathematics". Amodei said he believes AI for biology is on a similar trend.
The Verge described the result as "an early test of Claude's usefulness for science as the company prepares to go public". TechCrunch noted the lab was established in the spring, and that a discovery within a few months of opening would be notable for speed. Anthropic said humans remain in the loop and have not let Claude run autonomously in the lab without oversight.
ANALYSIS The announcement pairs a concrete but unvalidated scientific claim with a broader narrative about AI-driven research acceleration, timed as Anthropic approaches an IPO. The external scientific skepticism reported by Bloomberg and TechCrunch creates a gap between Anthropic's framing and independent assessment that will take time to resolve.