Anthropic Usage Policy Update June 2024

Anthropic has updated its Acceptable Use Policy to a new Usage Policy, effective June 6, 2024. These changes aim to clarify permissible applications of Anthropic's products and services to ensure responsible use as model capabilities and regulatory landscapes evolve.

Universal Usage Standards

Anthropic has consolidated its previous "Prohibited Uses" and "Prohibited Business Cases" sections into a single set of guidelines called "Universal Usage Standards." This consolidation ensures that the same set of rules applies uniformly to both individual consumers and business users.

Election Integrity and Misinformation

Anthropic's updated policy explicitly prohibits the use of its products for political lobbying, campaigning, and the dissemination of misinformation.

Key restrictions include:

• Political Advocacy: Products cannot be used to promote or advocate for a specific candidate, party, issue, or position.
• Political Activities: The use of models for soliciting votes or financial contributions is prohibited.
• Election Interference: The policy forbids using products to interfere with the election process, including targeting voting machines or obstructing the counting and certification of votes.
• Misinformation: The generation or participation in campaigns that spread false or misleading information regarding candidate information, election laws, and related topics is banned.

High-Risk Use Cases and Safety Requirements

Organizations using Anthropic's API for "high-risk use cases" are now required to implement additional safety measures. High-risk use cases are defined as integrations that provide information or analysis affecting healthcare decisions and legal guidance, where decisions may have significant consequences for individuals and require specialized expertise.

Expanded Access for Minors and Law Enforcement

While the Consumer Terms of Service generally prohibit use by individuals under 18, Anthropic now allows organizations to integrate the API into products designed for minors (such as tutoring or test preparation) provided they:

1. Implement specific safety features.
2. Disclose to users that the product leverages an AI system.

Additionally, Anthropic has expanded the list of countries whose law enforcement authorities may use its products for a limited set of use cases, specifically document summarization and call center support.

Privacy and Biometric Protections

Anthropic has strengthened its privacy protections by explicitly forbidding the use of its products for the following activities:

• Biometric Analysis: Analyzing biometric data to infer characteristics such as religious beliefs or race.
• Recognition Systems: Building recognition systems or techniques to infer emotions for use cases such as interrogation.
• Government Censorship: Analyzing or identifying specific content to censor on behalf of a government organization in any country where the products are permitted for use.