[DigitalToday reporter Chi-gyu Hwang] Anthropic has unveiled a new interface standard, the Model Hardware Standard (MHS), to help AI agents more easily operate machines and communicate with equipment.

CNBC reported on Aug. 27 that Anthropic appears to be seeking to expand the scope of AI applications to physical equipment through the standard.

MHS is designed to work on any device with a programmable interface. That includes scientific research equipment and advanced manufacturing facilities. Anthropic compared MHS to a USB-C cable that standardises how information is delivered between devices.

MHS focuses on reducing the time it takes companies to install and integrate equipment. It is also not dependent on a specific AI model. Users can connect not only to Claude-family models but also to other AI models.

Elizabeth Kelly (엘리자베스 켈리), head of beneficial deployments at Anthropic, said the company developed MHS to show AI potential in science and that it would also have major benefits for businesses and industry.

The announcement shows Anthropic is expanding its footprint into hardware. Anthropic is also building a silicon team to design custom chips for its own models. It recently hired hardware executive Caitlin Kalinowski, who previously worked at OpenAI, Meta and Apple.

MHS will first be offered as a research preview to some organisations in science, robotics and manufacturing. Anthropic plans to later release the standard as open source. Anthropic earlier released as open source the Model Context Protocol in 2024, designed to make it easier to connect AI agents to data sources.