Anthropic has established a wet lab in the San Francisco Bay Area as it expands its artificial-intelligence work into physical biology and drug science. The company says it wants to help unlock treatments for rare diseases and pursue preclinical work in areas traditional drugmakers consider financially unattractive. Its life-sciences chief, Eric Kauderer-Abrams, confirmed that Anthropic conducts some experiments in its own facilities and works with external partners, although a spokesperson said the lab is not specifically for drug discovery. The company is also exploring Claude-directed robotic systems, while saying human oversight remains essential for safety. Anthropic’s efforts come as its researchers and executives have publicly warned about catastrophic AI risks, including potential biological-weapons misuse, creating a need to balance faster medical development with deployment safeguards. The company has launched Claude Science, acquired Coefficient Bio for about $400 million in stock according to media reports, added Novartis CEO Vas Narasimhan to its board, and introduced the Model Hardware Standard to help AI operate laboratory equipment. Anthropic is targeting difficult or previously "undruggable" conditions, including potential work involving bispecific and trispecific antibodies, but has not disclosed specific diseases or progress and is not running clinical trials for now.