Police Agencies Use Veritone Track AI to Bypass Facial Recognition Bans
US law enforcement agencies, including the Department of Justice and Department of Homeland Security, have deployed Veritone's Track AI to identify and locate people in surveillance video by analyzing body attributes, gait, and clothing rather than faces. The tool reportedly circumvents facial recognition bans while raising similar privacy concerns, and critics warn of potential misuse for monitoring vulnerable populations.
What happened
Veritone has developed an AI model called Track that identifies and tracks individuals in video footage by analyzing body size, gender, hair color, clothing, accessories, and gait. The tool is designed to circumvent legal restrictions on facial recognition by using non-biometric attributes instead.
Track is used by 400 customers including state and local police departments across the US. The Department of Justice began using Track for criminal investigations in August 2024, and the tool is also used by the Department of Homeland Security and the Department of Defense.
Reported impact
The American Civil Liberties Union has expressed concern that Track raises similar privacy issues to facial recognition while introducing new ones. The organization warned about the tool's potential use for monitoring protesters, immigrants, and students, particularly given the stated focus on expanding federal monitoring capabilities.
Index analysis
Track currently operates on recorded video and can assemble timelines tracking individuals across locations and video feeds. The tool can access video from police body cameras, drones, and public sources, and is available through Amazon and Microsoft cloud platforms. Veritone indicated the technology will expand to process live video feeds within a year.
Governance controls
Veritone has stated that Track does not currently allow users to search specifically by skin color, though it uses skin tone as one of several attributes the algorithm employs to distinguish individuals. The departments using Track have not disclosed specific policies governing its deployment or use cases.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.