Usage Policy
Anthropic's rules for what Claude may be used for — the document your deployment is measured against, regardless of which plan or cloud you are on.
Last verified August 21, 2026 · 1 source
The Usage Policy defines what Claude may and may not be used for. It applies across every surface — the Claude apps, Claude Code, the API, and Claude accessed through Bedrock, Google Cloud or Foundry.
Why this reference does not summarise the prohibitions
A paraphrase of a policy is not the policy. Categories get added, wording gets tightened, and the version that governs you is the current one on Anthropic’s site — not a copy on a third-party wiki that was accurate in August 2026.
Read it at the source: anthropic.com/legal/aup.
What is worth saying here
It applies to what your users do, not just what you do. If you build a product on Claude, the policy reaches through your application to its users. That is a design constraint, not a legal footnote — it usually means you need your own terms and your own moderation.
Agentic use raises the stakes. A chat assistant produces text. An agent with computer use, file access and credentials takes actions with real effects. The policy is the same; the consequences of getting it wrong are not.
Enterprise controls exist to make compliance auditable. Audit logs, role-based access and the Compliance API on Enterprise are what turn “we comply” into something you can demonstrate.
Related documents
- Constitutional AI — how Claude is trained to behave
- Responsible Scaling Policy — how capability is gated
- System cards — what testing found before release
Sources
- 01Anthropic Usage PolicyOfficialanthropic.com
Page changelog
- VerifiedConfirmed as the governing acceptable-use document across surfaces.