Anthropic Launches Model Hardware Standard for AI Robotics

Anthropic launched a research preview of the Model Hardware Standard (MHS) on Thursday, introducing a model-agnostic software framework that allows AI agents to operate and coordinate physical machinery. Developed in collaboration with the HHMI Janelia Research Campus, MHS uses a standardized driver layer to connect large language models to devices with programmable interfaces, such as robotic arms, microscopes, and liquid handlers.

The standard aims to reduce hardware integration time from weeks or months to hours or minutes, facilitating autonomous, round-the-clock workflows for complex tasks including drug discovery and quantum computer laser calibration. By storing technical specifications within the framework, MHS allows AI assistants like Claude to operate equipment without relying on paper manuals or specialized human expertise. In one demonstration, a Genentech scientist used Claude to autonomously execute an experiment based on a PDF design.

Anthropic is currently limiting access to a select group of partners, including Amazon Web Services, Danaher, Doosan Robotics, Hugging Face, and Raspberry Pi. The company is using this preview period to develop a physical safety roadmap and safety evaluations to prevent misuse and address security concerns regarding authentication and physical hazards. Anthropic plans to open-source the standard and release its research findings and safety guidance once these evaluations are complete.