Anthropic’s biolab made a discovery it’s comparing to Crispr

Anthropic announced that its AI assistant Claude autonomously discovered a new enzyme system in bacteriophages after scanning a large database of DNA sequences. The company says nearly 950 Claude agents processed roughly 210 million tokens over a 21-hour period before one agent flagged an unusual repeating pattern for human review. Subsequent analysis and laboratory work conducted by Anthropic’s team reportedly confirmed the presence of an enzyme system that had not been previously characterized.

Anthropic characterized the finding as the first result from its newly launched wet lab and emphasized that human involvement was limited to the initial prompt to Claude and the experimental follow-up. The firm compared the newly identified system to the molecular machinery behind Crispr, though it also acknowledged that it is still working to determine the enzyme system’s function and potential applications.

The company framed the early disclosure as a way to demonstrate Claude’s capabilities and to provide the scientific community with insight into its work. At the same time, Anthropic’s announcement cautioned that it is unclear whether the discovery will have practical uses or be as transformative as Crispr, and the claim has been described within the reporting as premature.

The disclosure arrives amid a broader trend of AI companies applying large models to scientific research and laboratory workflows. Anthropic’s move also comes as the company prepares for a public offering and seeks to attract researchers to its lab and expand into areas such as drug discovery. Observers note that other AI firms are similarly testing model-driven approaches in scientific domains, raising questions about norms, validation, and the role of models in generating reproducible scientific results.