Anthropic has spent years warning about the risks of increasingly capable AI. Now it is testing what those systems can do inside a real biology lab.
The Claude maker has established a wet lab in the San Francisco Bay Area where it can conduct physical biological experiments, expanding its life sciences work beyond computer-based research. Eric Kauderer-Abrams, Anthropic’s head of life sciences, confirmed the facility to Reuters and said the company conducts some experiments internally while relying on outside partners for others.
Though Anthropic subsequently clarified to Reuters that the laboratory isn’t focused specifically on drug discovery, it declined to elaborate on the exact scope of the facility’s activities.
This initiative aligns with the company’s expanding commitment to pharmaceutical research, where it aims to tackle biological challenges and conditions that commercial drugmakers typically deem unprofitable.
From Claude to physical experiments
Anthropic is also exploring how Claude can work with laboratory equipment and robotic systems. People familiar with the effort told Reuters that the company wants AI to direct robotic units to perform experiments with limited human intervention. Kauderer-Abrams said the automation effort remains in its early stages.
“We’re in the very early innings of using AI to automate the execution of lab work,” he said in a Reuters report.
The company launched its Model Hardware Standard in August, an effort aimed at helping AI systems interact with laboratory equipment. Anthropic has also published research on protein design and biological modeling and launched a Life Sciences Verification Program for vetted researchers.
A bigger bet on biology
Anthropic’s interest in physical research follows its acquisition of Coefficient Bio, an AI biotech startup, and the appointment of Novartis CEO Vas Narasimhan to its board.
The company is also hiring for roles involving laboratory operations and protein and nucleic acid characterization. Kauderer-Abrams said life sciences is already one of Anthropic’s largest areas of investment by personnel and resources.
For Anthropic, owning laboratory capacity could shorten the distance between an AI-generated hypothesis and the experiment needed to test it. That could give the company firsthand experience with where AI performs well in biology, and where computer predictions break down.
More must-read AI coverage
- SS&C Intralinks DealCentre AI vs. Datasite: Which platform is built for the future of dealmaking?
- SS&C Intralinks FundCentre AI vs. Juniper Square: Which platform better supports modern private markets fund managers?
- Why Data, Not Models, Determines AI Success
- The Rise of the AI-Native Factory: How Physical AI Is Transforming Manufacturing
The safety problem
The expansion also creates an unusual tension for Anthropic. The company has recently warned about severe AI risks, including potential biological misuse, while simultaneously developing systems that can assist with biological research.
Anthropic has warned about the possibility of advanced models assisting with harmful biological work while simultaneously developing systems intended to help researchers analyze proteins, model biological processes, and interact with laboratory tools.
The company says human oversight remains important and has placed restrictions around access to some life sciences capabilities. But moving AI closer to physical experimentation raises the stakes because model outputs can increasingly be connected to real equipment and real biological processes.
That does not mean Claude is independently running complex research programs today. Anthropic says its automation work is still early, and the company has not disclosed the full scope of experiments taking place inside the lab.
What this means for users
Most people will not interact directly with Anthropic’s biology lab. The potential consumer impact is further downstream: if AI-assisted research eventually produces useful treatments, it could help researchers investigate diseases that have received less attention because of the cost or difficulty involved.
But that outcome is far from guaranteed. Anthropic has not disclosed specific diseases being pursued through its physical biology work, and it is not currently running clinical trials.
The more immediate shift is in what AI companies are trying to prove. Models such as Claude are moving beyond analyzing scientific papers and generating hypotheses toward interacting with the physical systems needed to test those ideas.
For Anthropic, the real experiment is no longer just whether Claude can reason about biology. It is whether those ideas still hold up once they reach the lab bench.
Other news: HP and OpenAI are partnering to bring agentic AI into enterprise workflows, with HP planning to integrate OpenAI models and agents across its hardware, services, and workplace tools.