Anthropic Usage Policy 2026-11-12

Citation

Anthropic. “Usage Policy” (also called the Acceptable Use Policy, AUP), policy web page, effective 2026-11-12. https://www.anthropic.com/legal/aup. Read 2026-10-15.

Bottom line

• The Anthropic Usage Policy has three tiers: Universal Usage Standards, High-risk Use Case Requirements and Additional Use Case Guidelines.
• It requires a “Qualified human-in-the-loop” and disclosure for advice and decisions in 11 “High-risk Areas”. These include credit, housing, employment, public benefits and legal status.
• It gives written exceptions for authorized security research. It also offers verification programs for “adjusted safeguards” in cyber and life sciences work.

Background

The policy applies to “anyone who submits inputs to Anthropic’s products or services”. This includes “customers accessing Claude through cloud providers”. The page shows the version “Effective November 12, 2026”, with a link to a “Previous Version”.

Key claims

Introduction:

• “Anthropic’s Safeguards Team implements detection and monitoring to enforce our Usage Policy. If we suspect that you may have violated our Usage Policy, we may warn you or throttle, limit, suspend, or terminate your access to our products and services.”
• “we may offer programs where you can apply to be granted access to our products and services with adjusted safeguards, such as our Cyber Verification Program and Life Sciences Verification Program”
• “Anthropic may enter into contracts with certain governmental customers that tailor use restrictions to that customer’s public mission and legal authorities”

Universal Usage Standards:

• “Do Not Develop or Design Weapons”: “Design, synthesize, produce, acquire, operate, test, deploy, or plan the deployment of biological, chemical, radiological, or nuclear weapons”.
• “Do Not Compromise Computer or Network Systems Without Authorization”: “Create or distribute malware, ransomware, or other types of malicious code”.
• The same section “does not prohibit security research, testing, or tool development conducted on systems you own or operate, with the authorization of the system owner”.
• “Do Not Compromise Critical Infrastructure”: “Facilitate the destruction or disruption of critical infrastructure such as power grids, water treatment facilities, medical devices, telecommunication networks”.
• “Do Not Use for Surveillance, Censorship, or Prohibited Law Enforcement or Criminal Justice Purposes”: “Target or track a person or their property without that person’s consent by monitoring their physical location, biometrics, movements, online activities, communications, or associations”.
• The same section prohibits use to “Make or suggest decisions in law enforcement or criminal justice processes”. It allows “authorized security research, legal research and analysis by law enforcement agencies or courts, or journalism”.
• “Do Not Compromise Privacy or Identity Rights”: “Misuse, collect, solicit, or gain access without permission to private information such as non-public contact details, health data, biometric or neural data (including facial recognition)”.
• “Do Not Create or Spread Disinformation”: “Impersonate real individuals or entities, or create fake personas or accounts, to falsely attribute content or mislead audiences about its origin”.
• “Do Not Engage in Fraudulent, Abusive, or Predatory Practices”: “Undertake, facilitate, or generate content for fraudulent activities, schemes, scams, extortion, phishing”.
• “Do Not Compromise Children’s Safety”: “Create, distribute, promote, or facilitate the creation of child sexual abuse material (“CSAM”), including AI-generated CSAM”.
• Anthropic reports CSAM: “we will report our findings to relevant authorities.”
• “Do Not Undermine Democratic Processes”: “Generate or disseminate false or misleading information in political and electoral contexts, including about candidates, parties, policies, voting procedures, or election security”.
• “Do Not Incite Violence or Hateful Behavior”: “Incite, facilitate, or promote violent extremism, terrorism, or hateful behavior”.
• “Do Not Engage in Cruel, Abusive, or Psychologically Harmful Conduct”: “Coordinate the harassment or intimidation of an individual or group”.

High-risk Use Case Requirements:

• “we require that a qualified person remain accountable for what is delivered or decided, and that the impacted individual knows AI was involved in the advice or decision.”
• Requirements, “Qualified human-in-the-loop for advice and decisions”: “A qualified person must meaningfully review a High-risk AI Recommendation, with the authority to change or modify it, before it is delivered”.
• Requirements, “Disclosure for advice and decisions”: the individual “must be clearly told that AI was used to produce it.”
• “High-risk AI Recommendations” lists 11 “High-risk Areas”:
  • Legal, Medical, Finance, Credit, Insurance, Housing
  • Employment, Education and credentials, Healthcare access
  • Public benefits and services, Legal status and adjudication
• “High-risk Physical Actions” adds requirements for hardware that the model controls, for example “Control hazardous energy or materials”.
• “Favorable Decisions”: “human-in-the-loop review is not required before acting on a High-risk AI Recommendation wholly favorable to the individual”. Examples are “Granting or continuing a public benefit or service.”

Evidence and method

This is a policy document, not research. It states rules. It gives no data on how often Anthropic enforces them.

Limits

• Until that date, the “Previous Version” applies.
• The download did not capture the text of the “Additional Use Case Guidelines” tab. That tab covers chatbots, products for minors, agentic use and software directories.
• The policy does not give enforcement data, appeal steps or time limits.
• The policy names CBRN (chemical, biological, radiological and nuclear) weapons in one bullet. It does not use the acronym.