Archive · 28 Aug 2026
Anthropic Introduces Standard for AI-Controlled Physical Equipment
Anthropic is testing a system called the Model Hardware Standard, or MHS.
It is designed to help AI agents control physical machines.
These machines can include microscopes, robotic arms, liquid handlers and other laboratory or manufacturing equipment.
MHS gives different devices a common way to communicate, somewhat like USB-C helps devices connect.
An AI could use it to coordinate several machines during an experiment or production task.
Anthropic says this could support drug discovery, manufacturing and specialised jobs such as quantum-computer laser calibration.
However, the AI may not always understand physical problems, such as bubbles in liquid-handling equipment.
The system is currently being tested by selected organisations, and Anthropic plans to make it open source after further safety evaluations.
Anthropic introduced a research preview of its Model Hardware Standard for AI agents.
The framework gives programmable devices a common interface for communicating with AI agents across networks.
It is designed to coordinate equipment such as microscopes, liquid handlers, robotic arms and plate readers.
Potential applications include drug discovery, laboratory operations, manufacturing and quantum-computer laser calibration.
The standard is being tested by selected organisations, is model-agnostic and may eventually be open-sourced.
- Who
- Anthropic, with selected organisations in science, robotics and manufacturing, is developing and testing the Model Hardware Standard.
- What
- Anthropic rolled out a research preview of a framework allowing AI agents to operate programmable physical devices.
- When
- The announcement was made on Thursday; one article was published on August 28, 2026.
- Where
- The framework is intended for scientific research, laboratories, advanced manufacturing and other network-connected physical environments.
- Why
- Anthropic says it could simplify hardware integration, coordinate multiple machines and enable more autonomous research and production workflows.
Potential Benefits
Safety and Reliability Concerns
Faster equipment integration
Potential Benefits
MHS could replace lengthy, device-by-device custom integrations with a common software interface. Anthropic cited a Carnegie Mellon integration that took about eight hours instead of weeks.
Safety and Reliability Concerns
The framework is still a research preview, and performance may depend on how well each device describes its capabilities, settings and safety limits.
Autonomous workflows
Potential Benefits
AI agents could coordinate multiple machines, monitor experiments, adjust parameters and handle repetitive research or manufacturing tasks with less human intervention.
Safety and Reliability Concerns
Anthropic’s tests showed that Claude initially responded incorrectly to bubbles during liquid handling by restarting the process, making the problem worse until a human explained the cause.
Open and model-agnostic standard
Potential Benefits
Anthropic says MHS is not limited to Claude and could eventually be open-sourced for adoption by device manufacturers and other AI systems.
Safety and Reliability Concerns
The standard is currently being shared only with selected organisations while safety evaluations and further development continue.
- Developer
- Anthropic
- Framework
- Model Hardware Standard (MHS)
- Current status
- Research preview available to a select group of organisations
- Supported equipment
- Programmable devices including microscopes, liquid handlers, robotic arms, plate readers and cameras
- Potential uses
- Drug discovery, laboratory operations, manufacturing and quantum-computer laser calibration
- Reported testing results
- Carnegie Mellon researchers said an integration took about eight hours instead of weeks and the experiment ran roughly three times faster
- Future plan
- Anthropic plans to open-source the standard after further development and safety evaluations
Sources
Anthropic launches Model Hardware Standard for AI agents
Anthropic wants AI agents to control physical machines with new hardware standard
Anthropic unveils ‘Model Hardware Standard’ to let AI agents control lab, manufacturing equipment
