Police AI adoption outpaces rules

Police departments are increasingly using AI to search body camera footage, analyze surveillance video, review digital evidence and draft reports, raising concerns among civil liberties advocates, legal scholars and policing experts. The tools can process large volumes of data quickly, but critics warn they may expand surveillance, embed hidden bias and make evidence harder to challenge in court.

Rules remain fragmented as agencies and state legislatures try to keep pace. California and Utah have enacted laws on generative AI in police report writing, including disclosure and accuracy safeguards, while other states have focused on related technologies such as facial recognition, drones and automated license plate readers. Mark43, a cloud software company serving more than 300 public safety agencies, offers AI tools for report drafting and case summaries, with audit logs and agency controls.

Experts are especially concerned about future “agentic policing” systems that could combine body-camera footage, camera networks and databases to generate leads or identify suspects. Misuse is already a concern: a former Costa Mesa, California, officer pleaded guilty in April to using law enforcement databases and Flock Safety cameras to monitor personal relationships, and at least 30 cities have ended or canceled license plate reader contracts since early 2025.

Some agencies are proceeding cautiously, while others are expanding deployments. Maui County approved a $1.7 million expansion of cameras and drones supported by AI, and Arkansas is developing a centralized criminal intelligence network. Experts recommend disclosure, human verification, independent audits, training and enforceable procurement standards to preserve accountability.