Anthropic has announced the opening of a research preview of the Model Hardware Standard (MHS), a shared specification designed to enable AI agents to safely operate physical devices. The announcement, made on August 27, 2026, marks a significant step toward standardizing safety protocols for embodied AI systems.
The MHS specification addresses one of the most challenging frontiers in AI development: creating reliable, safe interactions between AI agents and the physical world. Unlike software-only AI systems, physical AI agents must navigate real-world constraints, unexpected obstacles, and potential safety hazards that demand rigorous operational standards.
What is the Model Hardware Standard?
The MHS provides a unified framework for AI agents to interface with physical hardware in a safe, predictable manner. Key aspects of the specification include:
- Standardized safety protocols for robot movement and tool use
- Sensor validation frameworks ensuring reliable environmental awareness
- Emergency stop mechanisms with guaranteed response times
- Inter-agent communication standards for multi-device coordination
“We’re opening MHS to the research community because safety standards for physical AI can only improve through collaborative development,” said an Anthropic spokesperson. “This preview represents months of internal research and we want to work with the broader scientific community to refine these specifications.”
Initial Partners and Deployment
The research preview has been opened to a select group of scientific research labs and advanced manufacturers. Anthropic plans to gather feedback from these early adopters before making the specification more widely available.
The timing of this announcement is notable given the broader landscape of physical AI development. Multiple companies—including Tesla, Boston Dynamics, and various robotics startups—have been racing to deploy AI agents in real-world settings, from warehouses to construction sites. The lack of industry-wide safety standards has been a persistent concern among regulators and safety experts.
Industry Implications
The MHS could become a foundational standard for the emerging physical AI industry, similar to how safety standards shaped the automotive and aviation industries. By providing a common framework, Anthropic hopes to accelerate safe deployment of AI agents while reducing the compliance burden on manufacturers.
The specification arrives amid intensifying competition in the physical AI space. Google recently released Gemini Robotics ER 2, while various Chinese labs have been publishing work on vision-language-action models. Anthropic’s entry into this space through a standards-based approach represents a different strategic positioning—rather than building robots directly, the company is providing the safety infrastructure the industry needs.