AI camera providers could face liability for harm after police misidentification

The proposal would make a provider strictly liable for injury, death, or property damage caused by law enforcement acting because its tool misidentified a person or vehicle. The rule covers providers of systems that use AI to identify people or vehicles from camera images, when law enforcement acts on a mistaken identification. Police detention and certain responses to a person’s actions generally would not reduce or remove liability when they foreseeably follow from the mistaken identification. Liability may be reduced or removed when police action is justified by information independent of the mistake, or when the plaintiff’s known actions independently justify detention. The text does not state when this liability rule would begin.

The proposal would void agreement terms that waive, limit, or transfer a provider’s liability under this section. It applies to agreements involving a provider and law enforcement or a government unit. The restriction would not apply to an insurance policy purchased by the provider. This restriction would apply to agreements entered, changed, extended, or otherwise modified on or after the act’s effective date.