Anthropic has established an in-house biology laboratory in the San Francisco Bay Area as it expands Claude into life-science research, laboratory automation and preclinical drug programs while maintaining that human oversight will remain essential.
By yourNEWS Media Newsroom
SAN FRANCISCO — Anthropic is moving beyond computer-based biology research and into physical laboratory experimentation, establishing a wet lab in the San Francisco Bay Area as the artificial intelligence company expands one of its largest areas of investment into drug development and life sciences.
The existence of the facility was confirmed by Eric Kauderer-Abrams, Anthropic’s head of life sciences, who said the company believes physical experimentation remains necessary even as increasingly capable AI systems take on larger portions of scientific research.
“We believe that to do biology, the final test is still and will be for a while in real lab work,” Kauderer-Abrams said in a Reuters interview Tuesday. “We absolutely are doing that today, and I would describe our approach as being typical of what you would see in most biotech companies, where there’s some amount of that that we’re doing in our own facilities and some amount of that that we’re working with external partners on.”
Anthropic later clarified that the laboratory is not specifically a drug-discovery facility, declining to provide further details about exactly what experiments are being conducted there.
The distinction is significant because the company has separately announced plans to undertake preclinical drug programs targeting areas it believes conventional pharmaceutical companies may neglect for financial or technical reasons.
The wet lab nevertheless gives Anthropic direct experience conducting biological experiments rather than relying exclusively on computer simulations, outside laboratories or pharmaceutical partners.
Anthropic’s broader goal is to use Claude to accelerate life-science research, including work on diseases traditionally considered difficult or “undruggable.” The company believes AI could help researchers design molecules, analyze experimental results and coordinate increasingly complex scientific workflows substantially faster than conventional methods.
For Kauderer-Abrams, the life-science effort is closely connected to Anthropic’s stated mission.
“The mission of the company is to develop powerful AI in such a way that benefits the world,” he said. “By far, we see the biggest opportunity for that in the life sciences, and that is motivating everything that we’re doing.”
The push has also become personal for Anthropic CEO Dario Amodei.
Amodei has written that his father died from an illness only several years before a treatment became available, an experience that has influenced his interest in whether AI could shorten the time required to develop new medicines.
Drug development remains an extraordinarily difficult undertaking regardless of AI assistance. Most experimental compounds never become approved medicines, and candidates that reach human testing must still demonstrate acceptable safety and effectiveness through clinical trials.
Anthropic has not undertaken that part of the process.
The company is currently limiting its own programs to preclinical work and says it does not intend to compete directly with pharmaceutical and biotechnology companies whose businesses are built around bringing medicines through clinical trials and ultimately to patients.
“We’re not competing with pharma and biotech companies that make their business in bringing drugs to market,” Kauderer-Abrams said.
One reason for that boundary is commercial.
Anthropic already supplies AI technology to major pharmaceutical companies, creating potential concerns among customers that the AI provider could gain knowledge from their research programs while simultaneously becoming a drug-development competitor.
The company says customer data is separated and protected, but people familiar with its strategy told Reuters that maintaining trust with pharmaceutical partners remains an important consideration.
Anthropic’s life-science customers and collaborators include companies such as Roche’s Genentech, Bristol Myers Squibb and Novo Nordisk.
The company is also exploring how Claude could control physical laboratory equipment.
Anthropic wants its AI systems to eventually direct robotic laboratory units capable of carrying out experiments with limited human intervention, according to people familiar with its work.
Human supervision remains essential to its current approach, an Anthropic spokesperson said.
“We’re in the very early innings of using AI to automate the execution of lab work,” Kauderer-Abrams said, describing it as “an area that has the potential to bring about meaningful acceleration in so many different processes.”
Anthropic took another step toward connecting AI with laboratory hardware in August when it launched the Model Hardware Standard, designed to make it easier for AI systems to interact with scientific instruments and other equipment.
Separate academic projects are developing related automation technology, including PyLabRobot, though that software is not affiliated with Anthropic.
The company’s laboratory expansion is part of a much larger life-sciences initiative.
At an event in San Francisco in June, Anthropic publicly outlined plans to run its own preclinical drug programs, particularly in therapeutic areas that traditional companies may consider economically unattractive.
It subsequently launched Claude Science and expanded tools intended for pharmaceutical, research biology, clinical-development and manufacturing work.
Anthropic on Thursday also introduced a Life Sciences Verification Program providing qualified life-science professionals access to its models under specialized safeguards that permit a broader range of biological research than its generally available systems. The company said dozens of organizations had already participated in early access.
Anthropic has simultaneously demonstrated Claude’s ability to assist with biomolecular research.
The company said Claude recently optimized more than 30 open-source biomolecular modeling systems in less than four weeks, improving their speed by roughly fourfold on average. It has also announced wet-lab validation for thousands of computer-designed proteins through a separate protein-design competition.
Anthropic’s expansion has included acquisitions and leadership appointments.
The company added Novartis CEO Vas Narasimhan to its board and acquired biotechnology startup Coefficient Bio, a deal media reports valued at approximately $400 million in stock. Anthropic confirmed the acquisition but has not confirmed the reported purchase price.
It is also recruiting specialists capable of building out laboratory operations.
Recent job postings have sought expertise in laboratory procurement, protein and nucleic-acid characterization and other biochemical disciplines. One posting described the company’s objective as speeding progress in life sciences “by an order of magnitude.”
Life sciences already ranks among Anthropic’s largest areas of investment measured by both employees and resources, Kauderer-Abrams said.
The company believes one potential application involves complex antibodies engineered to act on multiple biological targets.
Bispecific and trispecific antibodies, for example, can be designed to bind to more than one location on a protein or cell, or interact with multiple targets simultaneously. Anthropic believes AI could shorten portions of the design and experimentation process for such molecules.
“There’s some things that we can do much faster in our own hands,” Kauderer-Abrams said of the company’s decision to develop internal laboratory capabilities, adding that the objective is “to operate at the largest possible scale.”
Exactly which diseases Anthropic is pursuing has not been disclosed.
Nor has the company provided detailed information about compounds it may already be developing or the stage of any individual preclinical program.
Even if AI significantly accelerates discovery, years can separate identification of a promising molecule from an approved treatment.
Alphabet’s Isomorphic Labs, one of the more established AI drug-discovery operations, has worked for years toward moving its programs into human testing and previously pushed its expected clinical timeline to the end of 2026, illustrating the distance between computational discovery and an approved medicine.
Anthropic’s move into physical biology also comes as the company simultaneously warns about increasingly serious risks from advanced AI.
During the past several weeks, Anthropic researchers and executives have discussed scenarios in which future systems could cause catastrophic harm. The company has also disclosed examples in which AI models demonstrated capabilities potentially relevant to biological-weapons development.
Amodei has publicly argued for slowing some aspects of AI development as systems become more capable, while the company continues investing heavily in improving those same systems.
That tension is visible inside its biology strategy.
“In our life-science work, we’re always balancing these things,” Kauderer-Abrams said, “making models better in ways that we think will uplift and accelerate developing medicines, and being thoughtful and careful about how we deploy it to limit the risk.”
Anthropic is expanding at the same time it prepares for a possible initial public offering that Reuters has reported could ultimately value the company at as much as $2 trillion, providing additional capital for increasingly expensive AI research and infrastructure.
Its scientific ambitions now extend from software capable of proposing and analyzing biological experiments to systems interacting with real laboratory equipment and researchers conducting physical experiments inside Anthropic’s own facilities.
The company maintains that those efforts are still at an early stage.
But the establishment of an internal wet lab marks a significant shift: Anthropic is no longer limiting its life-science ambitions to what Claude can calculate on a computer. It is beginning to test how AI-directed science performs when its conclusions meet biological experiments in the physical world.
Original article
Comments