# RAIBR03-BP02 Identify the severity of each potential harm
<a name="raibr03-bp02"></a>

 Risk severity estimates the magnitude of the negative on affected stakeholder groups if it were to occur. Severity also considers the reversibility of harm, recognizing that some types of harm may be permanent or difficult to remedy. 

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

## Implementation considerations
<a name="implementation-considerations-21"></a>

1.  Create standardized severity scale with clear impact levels. For example, establish a range from *low* (minimal, reversible impact) to *extreme* (substantial, long-lasting impact). Include specific criteria for each level to enable consistent evaluation. 

1.  Evaluate harm severity considering multiple factors. As an example, in medical AI systems, incorrect diagnoses might have *major* severity due to potential health impacts and difficulty in reversing treatment decisions. Consider immediate effects, long-term consequences and reversibility. 

1.  Document severity assessments with supporting evidence. For example, consider financial AI where incorrect investment advice might have a moderate or major severity estimate for impacted users, depending on the context. Include analysis of varying stakeholder impacts. 

## Resources
<a name="resources-20"></a>

 **Related documents:** 
+  [Learn how to assess the risk of AI systems](https://aws.amazon.com/blogs/machine-learning/learn-how-to-assess-risk-of-ai-systems/) 
+  [NIST Risk Management Framework](https://csrc.nist.gov/projects/risk-management/about-rmf) 
+  [ISO/IEC 42001:2023](https://www.iso.org/standard/42001) A.5.2 AI system impact assessment process