Loading...
Loading...
Governing invalidates options at the level of possibility—illegitimate actions are never generated, proposed, or optimised for. Filtering evaluates options after they have been generated, suppressing or modifying outputs without removing the underlying option from the decision space.
AI content moderation operates as filtering (blocking outputs after generation). Ex-ante AI governance would constrain the optimisation envelope itself, making certain outputs impossible rather than merely detectable.
GDS Section 5