Reuters found what Anthropic hadn’t been talking about: the company runs a wet lab in the San Francisco Bay Area — a place for real experiments on real molecules. Eric Kauderer-Abrams, who heads life sciences at Anthropic, confirmed it in an interview.
Why an AI company needs pipettes
“We believe that to do biology, the final test is still and will be for a while in real lab work,” Kauderer-Abrams said. Some of it happens in Anthropic’s own facilities, some with external partners — which, he argues, is how most biotech companies operate.
What the lab actually works on stays vague. A spokesperson later said it is not for drug discovery specifically, then declined to say what it is for. Two people familiar with the effort describe a goal beyond test tubes: Claude directing robotic units that carry out experiments with limited human intervention. Anthropic says human oversight remains essential for safety.
Life sciences is now one of the company’s largest areas by headcount and resources. The pieces around it fit: the roughly $400 million stock acquisition of Coefficient Bio, Vas Narasimhan on the board, the Claude Science software, and August’s Model Hardware Standard, which lets models operate lab equipment.
Where Anthropic draws the line
The company is not running clinical trials for now. Kauderer-Abrams partly puts that down to competitive concerns: “We’re not competing with pharma and biotech companies that make their business in bringing drugs to market.”
Underneath that sits a trust problem. Anthropic sells tools to large pharmaceutical companies while working on molecules itself. The two people say customers do worry the company could learn from their competing programmes, even with their data walled off.
For Dario Amodei the whole thing is personal anyway. A disease killed his father a few years before a cure arrived.
The timing is the awkward part
In the same fortnight this lab became public, Anthropic researchers warned that AI could end humanity, the company published examples where its systems could have contributed to biological weapons work, and Amodei called for slowing down. Meanwhile it is preparing an IPO at a rumoured $2 trillion valuation.
Kauderer-Abrams names the tension himself: “In our life-science work, we’re always balancing these things.” Make the models better so they speed up medicine — and be careful about how you ship them.
There is now a procedure for exactly that, published yesterday: the Life Sciences Verification Program, which lifts the biology safeguards for vetted labs. A company that has to keep its own safeguards straight is building a lab where it will need them itself. That isn’t a contradiction. It does explain why nobody volunteered the news.
Sources: