# RAIER01-BP01 Validate that release criteria still align with current industry standards
<a name="raier01-bp01"></a>

 At the start of a release evaluation, check that the release criteria and associated evaluation tests are still aligned with the current version of the AI system. Research and confirm that there are no new and relevant benchmarks or expectations that need to be included in the evaluation. 

 **Level of risk exposed if this best practice is not established:** Medium 

## Implementation considerations
<a name="implementation-considerations-79"></a>

1.  Compare your current release criteria against the actual system features and capabilities you plan to release, looking for gaps or mismatches. If your system includes capabilities that were not considered when you last updated your criteria, consider adding appropriate evaluation tests to cover these new features. This includes revisiting your risk and benefit assessment if necessary. 

1.  Stay up to date with new benchmarks, evaluation methods, or industry standards to see if there are new ways to test your system against your release criteria. 

1.  Consider new guidelines, updated regulations, or emerging compliance-aligned frameworks that might affect what you need to test before release. Consult with your legal team to assess relevant regulatory considerations. 

1.  Cross-check your evaluation datasets and test cases to make sure they still match the real-world scenarios where your system will be used. If your intended use cases have changed or expanded, you may need to update your evaluation approach to reflect these new applications. 

## Resources
<a name="resources-76"></a>

 **Related documents** 
+  [ISO/IEC 42001:2023 A.6.2.4 AI system verification and validation](https://www.iso.org/standard/42001) 

 **Related videos:** 
+  [AWS re:Invent 2024 - Responsible generative AI: Evaluation best practices and tools (AIM342)](https://www.youtube.com/watch?v=wuVpCc5a81Y) 

 **Related examples:** 
+  [awslabs](https://github.com/awslabs)/[agent-evaluation](https://github.com/awslabs/agent-evaluation) 
+  [aws-samples](https://github.com/aws-samples)/[rag-evaluation](https://github.com/aws-samples/rag-evaluation) 

 **Related tools** 
+  [Amazon Bedrock Evaluations](https://aws.amazon.com/bedrock/evaluations/) 
+  [Amazon SageMaker AI AI](https://aws.amazon.com/sagemaker/ai/?trk=bba24a8e-fec0-4c35-b7c7-d2e5e6b67eeb&sc_channel=ps&ef_id=CjwKCAjw2vTFBhAuEiwAFaScwgLGwsaX0LbsbBiFc16GhqyAGMIK79BPAbk_Bnl_-rlJVFq23-H2KRoCz5cQAvD_BwE:G:s&s_kwcid=AL!4422!3!724106169285!e!!g!!amazon%20sagemaker%20ai!19090032234!170269930766&gad_campaignid=19090032234&gbraid=0AAAAADjHtp97_-1psrdUeBS9kWnK-_Zmt&gclid=CjwKCAjw2vTFBhAuEiwAFaScwgLGwsaX0LbsbBiFc16GhqyAGMIK79BPAbk_Bnl_-rlJVFq23-H2KRoCz5cQAvD_BwE) 
+  [Amazon SageMaker AI Clarify](https://aws.amazon.com/sagemaker/ai/clarify/)