US police deploy AI virtual agents for undercover online surveillance
US police departments near the Mexico border have deployed Overwatch, an AI tool made by Massive Blue that creates realistic online personas to interact with and collect intelligence on suspects ranging from alleged human traffickers to political activists and college protesters. The technology reportedly had not led to any known arrests as of summer 2024 and is part of a broader trend of private companies selling secretive surveillance tools to law enforcement.
What happened
American police departments near the United States-Mexico border deployed Overwatch, an AI system made by Massive Blue that creates realistic online personas to interact with and collect intelligence on suspects. The personas are designed to infiltrate networks across social media, text messages, Telegram, Signal, and SMS.
Massive Blue created multiple AI personas with fabricated backstories, including a 36-year-old divorced woman interested in activism and baking, a 25-year-old from Dearborn Michigan, and a 14-year-old boy, among others labeled as 'child trafficking AI persona,' 'AI pimp persona,' 'college protestor,' and 'juveniles.'
Reported impact
As of summer 2024, the technology had not led to any known arrests. Police targeted suspects ranging from alleged violent sex criminals to vaguely defined 'radicalized' political activists and 'college protesters' with the AI agents.
The deployment raised concerns about surveillance of political protest activities, particularly regarding students protesting Israel's war in Gaza and other college protesters.
Index analysis
The use of AI personas to impersonate real people raises questions about appropriate law enforcement techniques and the potential for entrapment or manipulation of suspects through deceptive AI agents. The vague targeting of 'radicalized activists' and 'college protesters' alongside criminals suggests the system may be used against protected political speech.
Governance controls
Internal documents and contracts obtained through public records requests revealed the technology as secretive and unproven at the time of deployment. No regulatory approval or oversight framework for the use of such AI surveillance tools was evident from the reporting.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.