AI model watermarking changes agent behavior

This matters as watermarking can enforce AI safety by tracking and modifying agent behavior to prevent misuse.

Auto-scored: The event is a minor risk but represents substantive governance in AI oversight.