Anthropic has set up a wet lab in the San Francisco Bay Area, taking part of its expanding life-sciences operation from software and computer models into physical experiments.
Eric Kauderer-Abrams, the company's head of life sciences, confirmed the facility in an interview with Reuters. He said Anthropic now divides real laboratory work between its own facilities and external partners, an approach he compared with the mix used by many biotechnology companies.
The scope is narrower than the phrase “AI drug lab” might suggest. An Anthropic spokesperson told Reuters that the facility is not specifically for drug discovery and did not provide more detail. The company has discussed running its own early-stage programs in areas that established drugmakers may consider commercially unattractive, but the diseases it is targeting and the progress of that work remain unclear.
Anthropic is also drawing a line before human trials. Kauderer-Abrams said the company is focused on preclinical work and is not running clinical trials for now. He said Anthropic does not intend to compete with pharmaceutical and biotech companies whose business is bringing medicines to market.
More access to Claude, with more scrutiny
The physical lab arrives alongside a broader push to make Claude useful for professional biology. On Sept. 17, Anthropic launched a beta Life Sciences Verification Program for qualified teams and institutions. Standard grants give verified researchers access to models with less restrictive biology safeguards, while separate, project-specific grants are required for higher-risk work.
Anthropic says it reviews applicants' research credentials, security practices and ethical oversight, then monitors use against the scope each organization declared. That framework reflects a basic tension in the company's biology strategy: the same models that can help researchers analyze experiments or design candidates may also provide dangerous assistance if access is misused.
The company has been studying that risk in physical settings for years. Anthropic said a small 2024 wet-lab pilot involving eight participants found no evidence that Claude improved performance on basic laboratory protocols, though the study was too small to settle the question. Its new lab is a commercial research operation rather than that safety trial, but both efforts put Anthropic closer to the point where AI-generated ideas have to survive contact with actual biology.
Tech