Anthropic previews Model Hardware Standard for AI-controlled lab machines

MHS uses software drivers that translate between computer systems and individual machines, allowing equipment with different interfaces to communicate consistently. Users can enter machine information in natural language, and MHS creates reference material describing capabilities, adjustments, and safety restrictions. Agents can then discover compatible equipment across networks without needing separate software bridges for every device.

The standard supports microscopes, liquid handlers, robotic arms, and other equipment with programmable interfaces. Anthropic claims integration that once required weeks or months could take hours or minutes for MHS-compliant machines. During early physical trials, Claude was tested with laser alignment supported by camera-based observation; the model adjusted the laser, checked the image, and repeated the process while assessing each change.

Anthropic has shared MHS with manufacturers and research groups in biotechnology, robotics, and quantum computing. Early participants include Amazon Web Services, Automata, Danaher, Doosan Robotics, MBF Bioscience, QIAGEN, Tecan, and Universal Robots.

Anthropic acknowledges that AI agents still require expert supervision because language models have limits when reasoning about physical equipment. MHS currently cannot support machines lacking a programmable interface. The company says it plans to work with manufacturers to add drivers and expand compatibility, and its eventual open-source release will include safety test findings and implementation guidance.