Evaluating AI Providers' Frontier Safety Frameworks
Summary
A study assessing 12 frontier AI safety frameworks using 65 criteria across risk identification, analysis, treatment, and governance, finding low median scores and limited accountability.
Key quotes
Overall scores range from 34% (Anthropic) to 8% (Cohere), with a median of 18%.
The paper evaluates the effectiveness of safety frameworks published by AI companies following the 2024 AI Seoul Summit. It compares these frameworks against established risk management principles from high-risk industries.