Anthropic Usage Policy
Card Grid View — Anthropic (2026)
1. Policy Structure (Three Tiers)
- Universal Usage Standards
- Prohibited activities applying to ALL users
- Covers illegal content, harmful activities, fraud, and abuse
- Foundation of responsible AI use
- High-Risk Use Case Requirements
- Extra safeguards for domains like legal, healthcare, finance, employment
- Requires human oversight and AI disclosure
- Additional Use Case Guidelines
- Specific rules for consumer-facing chatbots and MCP servers
- Disclosure and transparency requirements
2. Universal Usage Standards
- Prohibited activities
- Creating illegal or harmful content
- Fraud, deception, and misleading practices
- Harassment, hate speech, and violence
- Unauthorized access or security violations
- Weapons development and surveillance abuse
- Scope
- Applies to all users of Anthropic's products
- Violations can result in access termination
3. High-Risk Use Cases
- What counts as high-risk
- Legal advice and interpretation
- Healthcare diagnosis and treatment recommendations
- Financial advice and investment decisions
- Employment decisions (hiring, firing, evaluation)
- Any area where AI decisions significantly affect individuals
- Key requirement
- Human-in-the-loop review before outputs are finalized
- Disclosure that AI was involved in generating the advice
4. Consumer-Facing Chatbots
- Mandatory disclosure
- All consumer-facing chatbots must disclose they are AI
- Disclosure required at the beginning of each session
- Users must know they are interacting with an AI system
- Scope
- Applies to ALL consumer-facing chatbots
- Regardless of whether they handle high-risk use cases
- Includes external-facing interactive AI agents
5. Human-in-the-Loop Requirements
- Why human oversight is required
- High-risk decisions need human accountability
- AI can make mistakes with serious consequences
- Humans must validate AI before outputs affect people
- What this means in practice
- Qualified human review before finalization
- Human can override AI recommendations
- AI is an assistive tool, not an autonomous decision-maker
- Operator maintains responsibility for outcomes
6. Transparency & Disclosure
- Disclosure obligations
- When AI outputs are presented directly to consumers
- Must clearly indicate AI-generated content
- No deception about human vs AI origin
- Why transparency matters
- Informed consent for AI interactions
- Prevents misleading users about AI capabilities
- Builds trust in AI systems
- Essential for accountability
7. Enforcement & Compliance
- Who the policy applies to
- All users of Anthropic's products and services
- Developers building on Anthropic's API
- MCP server operators listed in the Connector Directory
- Consequences of violation
- Anthropic can suspend or terminate access
- Policy may be updated; users must stay compliant
- Safeguard team monitors compliance