Anthropic opens AI-powered biology research lab

Anthropic PBC has opened a wet lab in the San Francisco Bay Area, and it plans to use robots running on its Claude models to automate some of the work there. Reuters reported the setup on Thursday. The company hasn’t said exactly what the lab will study, which leaves the bigger question hanging: is this a drug push, a biology platform, or both?

Anthropic did launch a separate effort in June to develop drugs for diseases the pharmaceutical industry tends to leave alone. But a company spokesperson told Reuters the new lab is not “for drug discovery specifically.” That sounds narrower on the surface, though it also leaves room for other biology projects. The lab may end up being a place where Anthropic tests how far its models can go when they’re not just chatting, but actually steering lab equipment.

That’s where the Model Hardware Standard comes in. Anthropic says MHS lets Claude control lab gear, spot mistakes in experiments, fix them, and write automation scripts. In practice, that matters because biology work often means chaining together several instruments in a strict order, and researchers usually have to script that by hand. If Claude can do that reliably, the tedious middle layer of lab work starts looking a lot less human.

Anthropic has already been poking at the science problem itself. Before Claude Mythos 5.1 was released earlier this month, the company tested it on high-affinity binders, the molecules used to attach medicine to harmful proteins. Anthropic says the model produced 12 candidate molecules and hit a 50% success rate, which it says is far above the usual 10% to 15% seen in protein design projects. The company also told Reuters it isn’t trying to compete with pharma and biotech firms, and said some lab work will go to external partners. That points to a familiar arrangement: Anthropic may be building the tools and the research engine, then handing the hard commercial slog to someone else.