Anthropic’s updated Usage Policy takes effect November 12 and will require a qualified operator to watch and be able to stop equipment that could injure someone when it acts on Claude’s output without a person approving each action.
The requirement covers high-risk physical actions, including mobile robots moving through spaces where people may be present and machinery that could injure someone if it malfunctions. Anthropic’s list also includes equipment that controls hazardous energy, acts on a person’s body, or runs an industrial process where a fault could cause harm.
A qualified person must be able to observe the equipment and stop it at any time. The equipment must stop or hold a safe state if that person intervenes or its connection to Anthropic’s services is lost. Limits such as speed or force must be enforced by the machine or a separate controller, rather than relying on Claude’s output to keep operation within bounds.
For high-risk advice and decisions, Anthropic defines a “qualified person” as someone with the training or experience to assess the output in that field and a license when the law requires one. The physical-action rule also calls for a qualified operator, but does not set a separate credential for that role.
The rule does not cover systems that only monitor equipment, or systems that produce plans and commands for a qualified person to review before use. Anthropic says its policy does not replace laws or certifications for machinery, vehicles, aviation or medical devices.
The updated Usage Policy applies to people using Anthropic products and services, including developers and end users of products that integrate Claude. It is a company rule for using Claude, not a general certification that a connected machine is safe.
Sources
- Anthropic’s October 8 policy update
- Updated Usage Policy, effective November 12, 2026
- Previous Usage Policy, effective September 15, 2025
- Photo by Nenad Stojković, uploaded to Wikimedia Commons by Shixart1985, under CC BY 2.0; image and license record.
