Published
Report 242 Research — AI Safety Policy

Summary

Analyzes how safety measurement itself can produce iatrogenic harm through the FLIM Level 4 framework. When the evaluation tool shares the same blind spots as the intervention being evaluated, measurement becomes self-reinforcing rather than corrective.

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