Differing Cyber Evals Undermine Claims of AI Model Improvements

Users criticized inconsistent cyber evaluations across labs and auditors because they create misleading comparisons, communication failures, and results that undermine claims of AI model improvements.

3 comments with sentiment. View comments.