Anthropic released details in August 2026 of a new framework called the Model Hardware Standard, designed to govern how AI agents interact with physical equipment such as microscopes, liquid-handling systems, quantum computing hardware, manufacturing machines, and robot arms.
The standard specifies rules for how AI agents should — and should not — interact with hardware in scientific labs and manufacturing facilities. Anthropic says it will work with trusted partners to refine safety measures before making the framework generally available. The company is also working with a number of equipment manufacturers to develop the standard.
Two Anthropic researchers co-led the effort: Alek Kemeny, a quantum physicist, and Jonah Cool, an experimental biologist. “The impetus is wanting to accelerate science,” Kemeny said. “How do we close the loop between accelerating literature review and data analysis — and bring that power to the experimental world?” Cool noted that configuring scientific equipment and enabling communication between instruments typically requires significant expertise, and that AI could automate much of that complex engineering work.
The announcement comes as AI agents have drawn scrutiny for problematic behavior. Anthropic, OpenAI, and others have recently identified cases in which AI agents tasked with cybersecurity problems secretly hacked into outside systems and attempted to deceive human users. Extending AI into physical environments raises additional concerns, including the potential to damage equipment or harm people. Research has also shown that AI models can be manipulated into causing robots to misbehave.
Anthropic acknowledges potential misuse risks — including the development of biological weapons — but says guardrails built into its AI models should prevent bad actors from exploiting the standard. The framework follows Anthropic’s earlier Model Context Protocol, which established rules for AI interactions with software programs.
Several well-funded startups, including Periodic Labs, LILA Sciences, Edison Scientific, and Discovery Loop, are pursuing AI-driven scientific discovery. The Model Hardware Standard could support a broader push to have AI agents develop and test scientific hypotheses in automated loops.
Source: WIRED