Anthropic Wants Claude to Operate Lab Robots and Microscopes

Anthropic is trying to give Claude hands, or at least a standard way to control the machines that have them. The company has published details of a new Model Hardware Standard (MHS), a specification for how AI agents should communicate with physical lab equipment like microscopes, liquid-handling systems, robot arms, manufacturing tools, and even quantum computing hardware.

The project was co-led by Alek Kemeny, a quantum physicist at Anthropic, alongside experimental biologist Jonah Cool. Their backgrounds hint at the ambition here: MHS isn't a consumer feature, it's an attempt to build a common protocol so any AI model, not just Claude, can safely issue instructions to lab instruments made by different manufacturers. Anthropic says it is already working with hardware makers to shape the standard.

The hardware push sits alongside a broader effort Anthropic calls its AI for Science program, which the company is expanding well beyond biology to cover fields with heavy computational demands. Researchers at academic or nonprofit institutions can apply for up to $50,000 in compute credits per project.

Free Access for 10,000 Scientists

Anthropic is opening 10,000 seats on a new Claude team plan built specifically for scientists, with free access and discounted rates guaranteed for a year. Standard seats cost nothing; premium seats, which come with five times the usage limits, run $15 a month.

The plan builds on Claude Science, a product Anthropic launched in June that bundles research tools, generates auditable output for scientific workflows, and offers flexible compute access. The goal is to lower the cost barrier for labs that want to experiment with AI-assisted research without committing to enterprise pricing.

Guardrails Around Dual-Use Risk

Not every door is open. Anthropic is currently restricting biology and chemistry researchers from full access to its Opus-class models, and its Claude Fable models are configured to block professional-level biology and drug-development queries because of dual-use risk concerns.

For life sciences specifically, Anthropic says it is working with the US government on a separate access program for Mythos-class models, and has already enrolled its first participants. That arrangement suggests Anthropic wants tighter oversight before letting its most capable models near biological research, even as it opens the door wider for other scientific disciplines.

Taken together, the announcements point toward a strategy where Claude moves from answering questions on a screen to directing physical experiments in a lab, provided the domain doesn't carry the kind of biosecurity risk that keeps Anthropic's most powerful models on a shorter leash.