Anthropic tests a new standard for Claude to work with factory and lab hardware

image via TNW | The heart of techThe Next Web

August 27, 2026, 8:15 PM

• •Anthropic introduced the Model Hardware Standard, a research-preview specification for describing hardware capabilities and safe operating limits to AI models.
• •An example use is a robot arm, where an MHS file can restrict movement speeds and joint angles for safer operation.
• •Anthropic said a Genentech scientist sent Claude a PDF describing an experiment, which the assistant then executed autonomously on hardware using MHS.
• •Amazon Web Services, Danaher, Hugging Face, and Raspberry Pi are among the early testers.
• •The article links the standard to EU Regulation 2023/1230, which will cover AI-based safety functions in machinery from 20 January 2027.
• •Anthropic plans to open-source the framework after the research preview ends.

Anthropic has released the Model Hardware Standard in research preview as a way to describe how hardware works and how AI systems may operate it safely. The company says the format can encode limits such as a robot arm’s allowed speeds and angles, and it demonstrated Claude executing a lab experiment on equipment carrying an MHS specification. Testers include Amazon Web Services, Danaher, Hugging Face, and Raspberry Pi, and Anthropic plans to open-source the framework after the preview. The article argues that Europe’s Machinery Regulation, which begins covering AI-based safety functions on 20 January 2027, could make such control specifications part of a regulated safety component.

Read original article