Police agencies near the United States-Mexico border are experimenting with a new kind of undercover tool: AI-generated online personas built to talk with people on social media, messaging apps, and other parts of the internet.
The product, called Overwatch, is sold by Massive Blue, a New York-based company. According to internal documents, contracts, and communications obtained by 404 Media through public records requests, the technology is being marketed for public safety work including border security, school safety, and human trafficking investigations.
What Overwatch Claims To Do
Massive Blue describes Overwatch as an “AI-powered force multiplier for public safety” that “deploys lifelike virtual agents, which infiltrate and engage criminal networks across various channels.” In practical terms, the documents describe AI-created online profiles that can interact with people over text messages, social media, Discord, and other messaging services.
The pitch is not simply that software watches public posts. The more striking part is the proposed interaction. A presentation obtained by 404 Media says the virtual personas can be used across the internet with the specific purpose of engaging suspects.
The documents say the system scans open social media channels for potential suspects. They do not explain how Massive Blue decides who should be treated as a potential suspect based on online activity.
That missing explanation matters because the listed targets are broad. The reporting describes use cases and personas connected to suspected drug traffickers, human traffickers, gun traffickers, “college protesters,” and “radicalized” political activists.
The Personas Go Beyond Generic Profiles
404 Media obtained a Massive Blue presentation showing examples of the AI characters. The personas were not presented as empty accounts. They included backstories, demographic details, interests, and communications capabilities that appear designed to make them more believable online.
One example was a “radicalized AI” “protest persona.” The presentation described that persona as a 36-year-old divorced woman who is lonely, has no children, and is interested in baking, activism, and “body positivity.”
Another was a “‘Honeypot’ AI Persona.” Its backstory described a 25-year-old from Dearborn, Michigan, whose parents emigrated from Yemen and who speaks the Sanaani dialect of Arabic. The presentation also said the persona uses various social media apps, is on Telegram and Signal, and has US and international SMS capabilities.
Other listed personas included:
- A 14-year-old boy “child trafficking AI persona”
- An “AI pimp persona”
- “college protestor”
- “external recruiter for protests”
- “escorts”
- “juveniles”
The range of personas is central to the controversy. A tool promoted for serious crimes is also being associated in materials with protest activity and other categories that are not described with the same specificity as trafficking or narcotics investigations.
Contracts Show Public Money Behind The Tool
The largest contract described in the reporting is with Pinal County, Arizona, which is between Tucson and Phoenix. Massive Blue signed a $360,000 contract with the county, funded through an anti-human trafficking grant from the Arizona Department of Public Safety.
A Pinal County purchasing division report says the county bought “24/7 monitoring of numerous web and social media platforms” and “development, deployment, monitoring, and reporting on a virtual task force of up to 50 AI personas across 3 investigative categories.”
Yuma County, in southwestern Arizona, also tried the company. It signed a $10,000 contract in 2023, but did not renew it. A spokesperson for the Yuma County Sheriff’s Office told 404 Media “it did not meet our needs.”
Despite the spending, the record described in the source does not show arrests from the technology. The technology had not led to any known arrests as of last summer, and the Pinal County Sheriff’s Office told 404 Media that Massive Blue had not been used for any arrests so far.
“Our investigations are still underway. Massive Blue is one component of support in these investigations, which are still active and ongoing. No arrests have been made yet,” Sam Salzwedel, Pinal County Sheriff's Office public information officer, told 404 Media.
Salzwedel also said Massive Blue has generated leads that detectives are pursuing. He said the office could not provide specifics about personas because the investigations are ongoing.
Officials Defend Secrecy While Critics Question The Target
Massive Blue cofounder Mike McGraw did not answer specific questions from 404 Media about how the product works, which police departments use it, or whether it had produced arrests. He said the company supports investigations and prosecutions of human traffickers.
“We are proud of the work we do to support the investigation and prosecution of human traffickers,” McGraw said.
McGraw also said the company could not disclose proprietary information because doing so could jeopardize investigations and put victims in danger.
Pinal County officials have also limited public detail. At public appropriations hearings about the contract, the sheriff’s office declined to explain the product to county council members in detail. Matthew Thomas, Pinal County Deputy Sheriff, said he “can’t get into great detail” and that doing so would “tip our hand to the bad guys.”
The Arizona Department of Public Safety said it understood Pinal County planned to use technology to help identify and solve human trafficking cases, and that this was what the department funded. It also said it was unaware of the specifics of Overwatch.
Privacy and civil liberties concerns are sharper because the personas are not limited in the documents to clearly defined violent crime targets. Dave Maass, who studies border surveillance technologies for the Electronic Frontier Foundation, told 404 Media that the idea of fake online identities for investigations is not new, but questioned what problem the tool is solving.
“The problem with all these things is that these are ill-defined problems. What problem are they actually trying to solve?”
Maass pointed specifically to personas involving escorts and college protesters, asking whether the effective result could be “violating protesters’ First Amendment rights?”
Why This Matters Now
The documents show how AI is moving into undercover policing not only as an analysis tool, but as an active participant in online conversations. That raises practical questions about accuracy, oversight, evidence, and how people are selected for engagement.
The source reporting also places the issue in a broader political setting. Concerns have gained urgency because the Trump administration has revoked the visas of hundreds of students, many of whom protested against Israel’s war in Gaza.
For now, the public record described by 404 Media leaves major questions unanswered. Police agencies are spending public money on AI personas. The company and agencies say secrecy is needed for investigations. Yet the available documents do not explain how suspects are identified, and officials say no arrests have been made yet.