Anthropic is moving further into the physical world with a new interface designed to help AI agents communicate with and operate programmable machines. Called the Model Hardware Standard (MHS), the system is intended to work across devices used in areas such as scientific research, robotics and advanced manufacturing. Anthropic compared it to USB-C, which provides a standardized way for different devices to exchange information.
The company said MHS could reduce the time and complexity involved in setting up and integrating hardware with AI systems. Unlike standards tied to a specific AI model, MHS is model agnostic, meaning organizations would not be limited to Anthropic’s Claude models. The standard is currently available to a limited group of organizations through a research preview, with Anthropic planning to open-source it so manufacturers across industries can adopt the technology.
The move puts Anthropic into closer competition with companies such as OpenAI and Amazon, which are also investing heavily in AI-powered hardware and physical systems. Anthropic is expanding its own hardware capabilities, including building a silicon team to develop custom chips for its AI models. Its latest initiative builds on the company’s open-source Model Context Protocol, launched in 2024 to make it easier for AI agents to connect with data sources, and signals a broader push to make AI capable of interacting with the physical world.
Related Readings:
