The Operator Ratio Problem

New laws are professionalizing remote AV operators — licensing them, logging their decisions, tying credentials to outcomes. That's the right instinct. But the legislation stops before it closes the gap it's trying to close.

What the new legislation gets right

In 2026, Texas, California, and pending federal legislation under the SELF DRIVE Act moved to require that a licensed, identifiable human operator be accountable for consequential vehicle decisions — including decisions the vehicle's AI system escalates to them in real time.

The Waymo school bus incidents made the failure mode visible: a vehicle correctly identified a legal obligation, escalated to a remote human for confirmation, received an incorrect answer, and proceeded to break the law. The algorithm didn't fail. The accountability architecture did.

The logbook infrastructure proposed under the SELF DRIVE Act addresses this directly. Every escalated decision is cryptographically signed by a licensed operator with a unique federal ID. If an incident occurs, investigators pull the logbook. The chain of command is auditable. A licensed human made the call, and the record proves it.

This is the right structure. A licensed decision-maker. An audit trail. Liability that attaches to an identifiable person. These are the conditions accountability requires.

The gap it doesn't close

The legislation professionalizes who is responsible. It does not bound how much any one person can be responsible for simultaneously.

A licensed operator monitoring 40 vehicles isn't making 40 decisions. They are applying one decision — with degraded attention, compressed context, and seconds per vehicle — across 40 situations they cannot meaningfully distinguish. The logbook records their name on each one. It does not record that they had 1.4 seconds and 34 other open sessions when the school bus question came in.

Existing regulatory frameworks recognize this problem in adjacent domains:

• Trucking. Hours-of-service rules limit driver duty time because sustained operation degrades judgment. Electronic logging devices enforce the limit. The ELD doesn't just record who drove — it enforces the conditions under which driving is legal.
• Aviation. Crew rest requirements and strict duty hour caps exist because fatigue-related decision failure is a known, measurable risk. The license is only valid under specified operating conditions.
• Hospital nursing. California's nurse-to-patient ratio law established that accountability without capacity bounds is not accountability — it is legal exposure dressed as oversight. A nurse responsible for 20 patients simultaneously is not responsible for any of them in any meaningful sense.

Remote AV operation is the same problem. The logbook establishes who is responsible. It does not establish whether that person was in any position to exercise the responsibility assigned to them.

A model ratio standard

A vehicle-to-operator ratio cap — tiered by operational context — is the structural complement to operator licensing. Without it, the licensing regime creates a new liability assignment without a corresponding oversight capacity.

Ratio limits should reflect decision complexity and consequence, not vehicle count alone. Geofencing and real-time context detection — already built into AV systems — can trigger automatic ratio reassignment as vehicles enter higher-risk zones.

The federal logbook infrastructure already proposed under the SELF DRIVE Act can record simultaneous session counts per operator. Ratio violations become auditable events, treated the same as decision errors. Companies operating above ratio ceilings during an incident bear strict liability — the ratio violation is itself evidence of negligent oversight architecture.

Model legislative language

For incorporation into state AV operator licensing frameworks or federal rulemaking:

A licensed remote operator may not maintain simultaneous active oversight responsibility for more than the number of autonomous vehicles permitted under the applicable operational context tier established by [agency]. Operational context shall be determined by real-time geofencing data logged to the National Automated Vehicle Safety Data Repository. Any incident occurring while an operator's active session count exceeded the applicable tier limit shall create a rebuttable presumption of inadequate supervision.