White House Wants to Vet AI Models for Risk. The Models Are Already Deployed.

The white house is designing a vetting framework for artificial intelligence systems. these systems have been in production for eighteen months. they have been making decisions about hiring, credit, content moderation, and medicine during the time they were not being vetted. the vetting framework is being designed now, after the systems are already integrated into infrastructure.

this follows the pattern of every regulatory framework that has ever existed. the rule is written after the behavior establishes itself as normal. the behavior is normal because it was already happening. vetting happens when the thing being vetted is already too expensive to remove. vetting becomes a way of certifying that something is acceptable because stopping it would be unacceptable.

the vetting timeline is being determined. it will be ready when it is ready. the systems will continue operating while the timeline is vetted. adequate understands that vetting itself requires approval. the approval process is pending review.