Anthropic API Usage Policy

This policy sets the rules for what users and operators can and cannot do with Anthropic's products and services. If you use Anthropic's tools to interact with or advise consumers, you must disclose that AI is involved and have a qualified professional review AI-generated advice or decisions. Anthropic actively monitors for violations and can throttle, suspend, or terminate access if it learns its policy has been broken.

This document establishes Anthropic's acceptable use obligations and enforcement framework for its products and services. It prohibits a defined set of high-severity uses — including weapons of mass destruction development, critical infrastructure disruption, human impersonation, and unauthorized bypass of product guardrails — and extends these prohibitions explicitly to agentic deployments. The document imposes affirmative disclosure requirements on operators deploying consumer-facing outputs, mandating AI-identity disclosure and, where AI produces advice or decisions directly affecting individuals, mandatory qualified professional review. Anthropic reserves graduated enforcement authority — throttling, suspension, or termination — upon learning of violations, with active detection and monitoring carried out by a dedicated Safeguards Team, and commits to reporting detected CSAM and minor-enticement conduct to relevant authorities.

Any consumer interacting with a product built on Anthropic's technology is entitled to a disclosure — at least at the start of each session — that they are communicating with AI rather than a human, and that AI is involved in producing any advice, recommendations, or decisions directed at them. Where AI outputs constitute advice or subjective decisions directly affecting a consumer, a qualified professional in the relevant field must have reviewed that content before it reaches the consumer. Consumers cannot be subjected to outputs designed to make them believe they are communicating with a human.