Anthropic has opened a preview of its Model Hardware Standard (MHS), a proposed system designed to let AI agents control programmable physical machines through a common set of rules.
The standard is intended for laboratories, factories, and other workplaces where different instruments need to work together without requiring custom software connections for every device.
MHS uses software drivers that translate instructions between computer systems and individual machines. This allows equipment with different interfaces to communicate in a consistent way.
The drivers rely on basic commands for tasks such as reading measurements or changing settings, giving AI agents a standard method for operating connected equipment.
The system can also store information about individual machines that may otherwise remain in technical manuals or depend on knowledge held by experienced laboratory staff. Users can provide this information through natural language. MHS then creates reference material describing a machine’s capabilities, available adjustments, and safety restrictions.
AI agents can use this information to discover compatible equipment across a network without requiring a separate software bridge for every machine they need to control.
The standard can support microscopes, liquid handlers, robotic arms, and other equipment that provides some form of programmable interface. Anthropic claims that hardware integration processes that previously took weeks or months could be reduced to hours or minutes when equipment supports MHS.
According to Anthropic , Claude was tested in physical experiments during early trials, including a laser alignment task supported by camera-based observation. Claude successfully adjusted the laser, checked the resulting image, and repeated the process while evaluating each change. MHS also allows agents to combine commands from multiple devices, making it possible to create automated workflows that would otherwise require coordination between several systems.
Anthropic said Claude interacts with experiments and hardware in an “exploratory manner, much as a scientist would” during physical testing. The model can coordinate several instruments through a single interface instead of requiring separate control software for each connected device. Anthropic has shared MHS with manufacturers and research organizations working in biotechnology, robotics and quantum computing.
The company intends to use the preview to establish safety checks before making the standard more widely available to developers and equipment manufacturers. Amazon Web Services, Automata, Danaher, Doosan Robotics, MBF Bioscience, QIAGEN, Tecan and Universal Robots are among the early participants.
However, Anthropic acknowledges that AI agents still require expert supervision because language models have clear limitations when reasoning about physical equipment. MHS also cannot currently work with equipment that lacks a programmable interface, leaving some laboratory and industrial machines outside its present scope.
Anthropic plans to work with manufacturers to develop additional drivers and expand compatibility with more devices and robotics platforms. The preview will also be used to develop new safety tests and practical guidelines for deploying AI agents around physical equipment.
Anthropic says the eventual open-source release of MHS will include findings from these tests along with detailed guidance for safer implementation.
For now, the project remains experimental, with its practical use still dependent on hardware design, software access, and human oversight.
Get the latest tech news, telecom insights, and product launches wherever you prefer.
Add ProPakistani to Preferred Sources and see more of our stories in Google Search and Top Stories.
Technology and Automotive Specialist covering the latest cars, smartphones, AI breakthroughs, and...
Shares
