Anthropic is operating a lab that conducts biology experiments

Anthropic has confirmed that it operates a wet biology lab in the Bay Area, giving its AI models a place to run physical experiments instead of staying trapped in text and predictions. The company said the setup is being used for real lab work, and Eric Kauderer-Abrams, Anthropic’s head of life sciences, told Reuters that “the final test” in biology is still hands-on lab work for the foreseeable future.

The lab sounds like a fairly standard biotech operation, at least by the company’s own description. Anthropic said it does some research there itself and also works with external partners. It would not say what the lab is specifically working on, except that the focus is fundamental biology rather than drug discovery.

That distinction matters because Anthropic has been trying hard not to look like it is taking a swing at its own customers. It has major pharma partners, including a newly announced deal with Novo Nordisk on joint drug discovery, and it has already been criticized for releasing products that seemed to brush up against what those customers do. The company also bought Coefficient Bio, a stealth AI biotech, in April, which makes the lab look less like a surprise and more like the next step in a path it had already started walking.

At the same time, the company is pushing harder into life sciences access. This week it launched a Life Sciences Verification Program, which gives vetted bio researchers access to its most powerful models. And it has published recent reports on protein design and bimolecular modeling as part of its pitch that AI can help with drug discovery, even if the wet lab itself is officially framed as more basic research.

All of this is happening while Anthropic’s own people are warning about AI in very dark terms. A researcher’s recent resignation, a senior alignment lead putting extinction odds above 10% within a decade, and Dario Amodei’s call for industry self-regulation all sit uneasily next to a company running biology experiments in the Bay Area. The tension is obvious: if AI is too dangerous to trust, why hand it lab coats and pipettes?