AIBRIEF
The framework targets any device with a programmable interface, from microscopes to quantum computer laser calibration, and is headed to open source only after partner led safety work.
Anthropic on Thursday rolled out a research preview of Model Hardware Standard (MHS), a framework designed to let AI agents operate lab and manufacturing instruments through one common interface.
MHS targets any device with a programmable interface: microscopes, robotic arms, and laser calibration on a quantum computer. The point is to make those instruments addressable by AI agents across a network, so a single agent can chain routine work (drug-discovery screens, machine tending, optical alignment) into a longer autonomous run, according to Anthropic via the Reuters wire.
Anthropic is positioning MHS as a general-purpose protocol rather than another instrument-specific SDK, a bet that only pays off if the rest of the industry agrees to use the same lingua franca. Labs and factories already have options: SiLA for lab automation, OPC UA for industrial control, LabVIEW, and a long tail of manufacturer SDKs. Whether an AI-vendor standard displaces any of them is the open question, and one the announcement does not answer.
Anthropic is sharing an early version with partners specifically to build out safety evaluations before releasing the standard as open source. No release date beyond "Thursday" for the research preview, no named partners, and no pricing are in the announcement, so the practical timeline is still TBD.