Anthropic Usage Policy
The rules for anyone building on Claude, including the deceptive-use section that covers passing generated text off as human work. Read it before building a rewriting tool on the API.
What it is
The Usage Policy governs anyone building on Claude. Its deception rules cover presenting generated output as human-authored where that misleads, academic dishonesty, and impersonation. For this subject the relevance is direct: a tool whose purpose is to make Claude output pass as human is outside the policy, and Anthropic's own watermark announcement notes that the mark does not change a user's rights under the terms.
When to use this entry
Cite it when a question is about building rewriting or detection tools on Claude.
Questions
Can I build a humanizer on the Claude API?
The Usage Policy restricts deceptive uses, including passing generated text off as human work. Read the deception section before building.