AWS Responsible AI Policy 2025-01-13
Citation
Amazon Web Services, Inc. “AWS Responsible AI Policy.” Policy web page, last updated 2025-01-13. https://aws.amazon.com/ai/responsible-ai/policy/. Read 2026-10-15.
Bottom line
- This is the AI-specific hard-limits document of Amazon Web Services (AWS). It has seven prohibitions for “AI/ML Services”, including third-party models on AWS.
- It requires “appropriate human oversight” for “consequential decisions”, for example credit, housing, hiring and government benefits.
- It adds to the AWS Acceptable Use Policy 2021-07-01. Enforcement follows that policy.
Background
The policy “applies to your use of artificial intelligence and machine learning Services, features, and functionality (including third-party models) that we provide”. It “supplements the AWS Acceptable Use Policy and AWS Service Terms.”
Key claims
“Prohibitions”. “You may not use, or facilitate or allow others to use, the AI/ML Services”:
- “for intentional disinformation or deception;”
- “to violate the privacy rights of others, including unlawful tracking, monitoring, and identification;”
- “to depict a person’s voice or likeness without their consent or other appropriate rights, including unauthorized impersonation and non-consensual sexual imagery;”
- “for harm or abuse of a minor, including grooming and child sexual exploitation;”
- “to harass, harm, or encourage the harm of individuals or specific groups;”
- “to intentionally circumvent safety filters and functionality or prompt models to act in a manner that violates our Policies;”
- “to perform a lethal function in a weapon without human authorization or control.”
“Responsible AI Requirements”:
- “If you use the AI/ML Services to make consequential decisions, you must evaluate the potential risks of your use case and implement appropriate human oversight, testing, and other use case-specific safeguards”.
- Consequential decisions include “access to critical benefits like housing or government benefits, opportunities like education, decisions to hire or terminate employees, or access to lending/credit”.
- “You agree to provide information about your intended uses of the AI/ML Services and compliance with this Policy upon request.”
- “AWS may investigate and enforce violations of this Policy as noted in the AWS Acceptable Use Policy.”
Evidence and method
This is a policy document, not research. It states rules. It gives no data on how often AWS enforces them.
Limits
- The weapons rule covers only “a lethal function in a weapon without human authorization or control”. The policy does not name CBRN (chemical, biological, radiological and nuclear) weapons, cyberattacks or critical infrastructure. The general AWS policy covers attacks on system security.
- The policy does not name elections or political campaigning.
- The policy does not name research exceptions.
- Third-party models on AWS also have their own policies. For example, Anthropic’s policy applies to “customers accessing Claude through cloud providers” (Anthropic Usage Policy 2026-11-12).