The Pacing Problem

Oversight takes time. Investigations, rulings, fines, and laws move at the speed of institutions. AI deployments move at the speed of software. The Pacing Problem is the structural delay between the two: the system reaches scale, and the constraint arrives long afterward, if it arrives at all.

Clearview AI shows the gap at full size. The company scraped billions of images from the public internet to build a face-search tool and sold it to police forces. By the time a journalist surfaced it publicly, the database had been growing for years. The first major regulatory fine landed years after that, and it reached only one country's residents. Through every year of investigation, the database kept growing and the app kept running.

The reason is that oversight is built for a slower world. Each regulator has authority over a slice: one state's biometric law, one country's data-protection rule, one agency's mandate. Each investigation runs on its own clock and stops at its own border. None has authority over the whole system, and the system does not pause while they work. The accountability that hits some failures all at once arrives here slowly, in pieces, after the fact.

This is not a story about lazy regulators. The rulings were often correct. The fines were real. The problem is timing. A finding that a deployment was unlawful, delivered after the deployment has already reached everyone it was going to reach, changes the record but not the outcome. The constraint exists; it just arrives after the moment it could have mattered.

For AI specifically, the pacing gap is widening, because models ship and update in weeks while oversight cycles run in years. A system can move through several generations during a single investigation. Any organization relying on external oversight to catch its failures is relying on a clock that runs slower than the one its system runs on. The gap between deployment and constraint is where the harm lives.