The mechanism of AI harm

2025

Open publication workspace · Sign in to read the full PDF.

AI-generated summary

1) Understanding the mechanisms of AI harm is crucial for developing effective strategies to prevent and mitigate its negative impacts.
2) This report analyzes AI incident data to identify six key mechanisms of harm: harm by design, AI misuse, attacks on AI systems, AI failures, failures of human oversight, and integration harm. It argues that a one-size-fits-all approach to harm prevention is insufficient and that diverse, sociotechnical strategies are required.
3) The report highlights that AI capabilities alone are not adequate predictors of harm, and that context, manner, and governance are stronger determinants. It also emphasizes the importance of comprehensive incident tracking for identifying and responding to AI risks.
Tags: AI harm, AI risks, AI governance, incident analysis, mitigation strategies

Check the original publication for accuracy and context.