Understanding AI Harms: An Overview

Heather Frase

Owen Daniels

August 11, 2023

As policymakers decide how best to regulate AI, they first need to grasp the different types of harm that various AI applications might cause at the individual, national, and even societal levels. To better understand AI harm, the blog presents some key components and characteristics.

Real-world harms caused by the use of AI technologies are widespread. Tracking and analyzing them improves our understanding of the variety of harms and the circumstances that lead to their occurrence once AI systems are deployed. This report presents a standardized conceptual framework for defining, tracking, classifying, and understanding harms caused by AI. It lays out the key elements required for the identification of AI harm, their basic relational structure, and definitions without imposing a single interpretation of AI harm. The brief concludes with an example of how to apply and customize the framework while keeping its modular structure.

On July 21, the White House announced voluntary commitments from seven AI firms to ensure safe, secure, and transparent AI. CSET’s research provides important context to this discussion.

Artificial intelligence systems are rapidly being deployed in all sectors of the economy, yet significant research has demonstrated that these systems can be vulnerable to a wide array of attacks. How different are these problems from more common cybersecurity vulnerabilities? What legal ambiguities do they create, and how can organizations ameliorate them? This report, produced in collaboration with the Program on Geopolitics, Technology, and Governance at the Stanford Cyber Policy Center, presents the recommendations of a July 2022 workshop of experts to help answer these questions.

How can we measure the reliability of machine learning systems? And do these measures really help us predict real world performance? A recent study by the Stanford Intelligent Systems Laboratory, supported by CSET funding, provides new evidence that models may perform well on certain reliability metrics while still being unreliable in other ways. This blog post summarizes the study’s results, which suggest that policymakers and regulators should not think of “reliability” or “robustness” as a single, easy-to-measure property of an AI system. Instead, AI reliability requirements will need to consider which facets of reliability matter most for any given use case, and how those facets can be evaluated.