Anthropic Updates Its Usage Policy: What Changes on 12 November

On 8 October Anthropic published a new version of its Usage Policy, the rules on what Claude can and cannot be used for. It applies from 12 November 2026 to all its products: the Claude app, Claude Code and the API.

The main changes

• Deceptive activity (new section). It brings together the rules on fake accounts and fabricated content, including hiding who is behind a message or amplifying it through fake accounts.
• Elections. The section is renamed “Do Not Undermine Democratic Processes”. It drops the blanket ban on personalised voter targeting and focuses on stopping Claude being used to deceive voters or disrupt elections.
• Weapons. It now states that software and components for weapons are also banned, including arming drones and other autonomous vehicles.
• Surveillance. It bans tracking people without consent, in real time or through analysis, and recommending arrests or charges. It allows uses such as fraud monitoring, content moderation and journalism.
• High-risk uses. Health, legal and financial decisions still need human review. And something new: if an agent takes physical actions through hardware, qualified operators are required.
• Abusive behaviour (new section). It bans sustained and needless abusive or cruel behaviour towards the model. Anthropic says this only covers extreme cases, not ordinary frustration or testing Claude.

What it means for you

• If you use Claude to study, work or code, nothing changes. The changes mostly target companies and organisations.
• If you build agents that act in the physical world (robots, drones, devices), review the high-risk section before November. Our agent safety guide explains how to add human review to important actions.
• If you use AI in political campaigns or communication, read the new democratic processes section: what matters now is not deceiving people.