Published
Report 224 Research — AI Safety Policy

Summary

Analysis of iatrogenic risks when frontier models achieve 0% ASR. Examines whether rapid safety improvement creates false confidence, measurement challenges, and structural blind spots. Connects to FLIM framework and polyhedral refusal geometry findings.

This research informs our commercial services. See how we can help →