Anthropic's new usage policy applies from 12 November 2026

Anthropic's new usage policy adds hardware controls from 12 November

8 October 2026: Anthropic has published a rewritten Usage Policy that takes effect on 12 November 2026. Business customers building on Claude get new conditions where the model drives physical equipment, a tighter ban on tracking people without consent, a consolidated rule against fake accounts and deceptive campaigns, and a clearer statement of who may use the service from outside supported regions.

The hardware section matters most for operational deployments. Where model output makes equipment act without human approval, the policy now requires a qualified person who can observe the equipment and stop it at any time. Operating limits must be enforced by the equipment or an independent controller, not the model, and the equipment must stop or hold a safe state when that person intervenes or when the connection to Anthropic's services is lost. Triggers include moving through shared space, applying force that could injure, controlling hazardous energy or materials, acting on the human body, running safety systems and industrial processes. Exclusions cover consumer devices within built-in safety limits, plans or code reviewed by a qualified person before execution, monitoring without control, and lab automation with non-hazardous materials in enclosed instruments.

The surveillance section bars using Claude to make or suggest law enforcement or criminal justice decisions, to track people's location, biometrics, movements, online activity or communications without consent, or to build surveillance tools. Consented analysis such as fraud detection, aggregated or anonymised analysis, content moderation, authorised security research and journalism remain allowed.

A new section on deceptive campaigns and artificial activity prohibits fake personas, accounts, media outlets or organisations, hidden sponsorship of influence campaigns and seeding AI or search sources with misleading content. Fake reviews and bots posing as humans are named examples, which reaches marketing and review-generation tools. Weapons rules now cover guidance and control software, and the elections section is narrowed to deceiving voters or disrupting elections.

On regions, the policy forbids giving access to Claude to end users in breach of the Supported Regions Policy. Anthropic's announcement adds that the restriction covers people physically in an unsupported region, entities incorporated or headquartered there, and entities majority-owned or controlled from there.

Companies that resell or embed Claude should check their terms with their own customers before 12 November. Our AI model category lists 112 products.

Sources

• Anthropic: 2026 Usage Policy update, 8 October 2026
• Anthropic: Usage Policy (effective 12 November 2026), 8 October 2026