Anthropic is taking its AI ambitions into the laboratory, opening a physical biology facility in the San Francisco Bay Area as the company works to push Claude beyond computer simulations and into real-world scientific experiments.
The move marks a new phase in Anthropic's growing focus on life sciences. The company has been working on ways for AI systems to assist with biological research, but its latest efforts involve physical laboratory equipment, robotic systems and experiments that take place outside a computer.
Anthropic's head of life sciences, Eric Kauderer-Abrams, confirmed the company's new wet lab in an interview with Reuters. He said the company is already conducting physical biology work both in its own facilities and through outside partners.
“We believe that to do biology, the final test is still and will be for a while in real lab work,” Kauderer-Abrams told Reuters.
The company's longer-term ambition goes further. Anthropic wants Claude to be able to direct laboratory robots and coordinate experiments with limited human intervention, potentially allowing researchers to run far more experiments without having to manually operate every piece of equipment.
That vision is already being tested.
In August, Anthropic introduced its Model Hardware Standard, a system designed to give AI agents a common way to communicate with laboratory instruments and other physical equipment. The system can connect Claude to devices including liquid handlers, robotic arms and plate readers, allowing the AI to coordinate multiple machines during an experiment.
In one demonstration with biotech company Genentech, Claude coordinated a protein assay involving several pieces of laboratory equipment. The AI could adjust experimental parameters based on the results and use the equipment to carry out the next stage of the experiment.
Anthropic says the eventual goal is an “autonomous discovery engine” in which scientists provide the high-level biological objective while AI systems handle much of the routine physical work involved in carrying out experiments.
That could include generating instructions for laboratory equipment, running experiments, analyzing results and adjusting the next experiment based on what happened in the previous one.
Targeting diseases that have been difficult to treat
Anthropic has said its life-sciences work is particularly interested in diseases that have traditionally received less attention from drug developers, including rare conditions and so-called “undruggable” diseases.
The company believes AI could help researchers investigate biological targets that have been considered too difficult to address with conventional drug-development approaches. One area it has pointed to is the development of complex antibodies capable of targeting multiple points on a protein or cell.
But Anthropic is not currently trying to take a drug all the way from the laboratory to a patient.
The company says its work will remain focused on the preclinical stage rather than clinical trials, meaning it will not initially compete directly with pharmaceutical and biotechnology companies whose businesses involve bringing drugs to market.
“We’re not competing with pharma and biotech companies that make their business in bringing drugs to market,” Kauderer-Abrams said.
That distinction is important because Anthropic already provides AI tools to major pharmaceutical companies, including Genentech, Bristol Myers Squibb and Novo Nordisk.
From AI assistant to laboratory operator
Anthropic's push into physical science also reflects a broader change in how the company sees AI being used.
Rather than simply asking Claude to summarize scientific papers or suggest an experimental approach, researchers could eventually give an AI system a biological objective and have it coordinate the physical experiment itself.
Anthropic's hardware system is designed to make that possible by creating a common interface between AI agents and machines that would normally require specialized programming to operate together.
The company says its early experiments have already shown that Claude can coordinate multiple devices, detect problems and adjust certain experimental parameters in real time. In one demonstration, an AI-controlled system detected bubbles during a laboratory procedure and suggested using a centrifuge to correct the problem before continuing.
Anthropic acknowledges that the technology is still at an early stage. Its own researchers say there is substantial work to be done before AI can reliably handle the complexities of biological experiments, particularly when working with living cells and other sensitive materials.
Human scientists will also remain involved. Anthropic says human oversight is essential, particularly as AI systems gain greater control over physical laboratory equipment.
For now, the company's new wet lab represents a relatively small step toward a much larger ambition: using increasingly capable AI not just to think about science, but to actually perform parts of it.
And Anthropic is betting that the next major leap in AI-powered drug research may happen not inside a computer, but at the lab bench.
