Flock Has a Powerful New AI Tool for Police. We Got Its Code
Flock Safety has developed "OS Investigate," an AI tool that can identify drivers and track vehicles by movement patterns alone, contradicting its long-held public claim that its technology "cannot recognize, identify, or track individuals." The system features 69 prewritten prompts and access to 45 data tools, drawing from plate scans, arrest records, 911 dispatch logs, ballistics results, and commercial databases containing Social Security numbers, phone numbers, and relatives' information. Ma
Analysis
TL;DR
- Flock Safety has developed "OS Investigate," an AI tool that can identify drivers and track vehicles by movement patterns alone, contradicting its long-held public claim that its technology "cannot recognize, identify, or track individuals."
- The system features 69 prewritten prompts and access to 45 data tools, drawing from plate scans, arrest records, 911 dispatch logs, ballistics results, and commercial databases containing Social Security numbers, phone numbers, and relatives' information.
- Many prompts enable pattern-based surveillance without requiring a plate, name, or known crime—officers can search by location, time window, and behavior alone, such as vehicles visiting multiple banks or gas stations at odd hours.
- The tool can generate detailed dossiers on individuals, including associates ranked by co-occurrence at cameras, and perform one-command "workups" that return vehicles, prior suspect listings, relatives, phone numbers, and online accounts.
- Privacy experts and former law enforcement officers have criticized the system as unconstitutional warrantless surveillance, while Flock's CEO has described investigators as becoming "addicted" to the tool's capabilities.
Why It Matters
This development represents a significant escalation in AI-powered mass surveillance, shifting from targeted license plate lookups to proactive behavioral pattern analysis that can cast suspicion on ordinary citizens based solely on their movement. For AI practitioners and researchers, it raises urgent questions about the ethical deployment of agentic AI systems in law enforcement and the dangers of preprogrammed prompts that automate suspicion. The case also highlights how commercial surveillance companies can rapidly expand capabilities beyond their original public commitments, with profound implications for civil liberties and constitutional law.
Technical Details
- OS Investigate (formerly Nightshift) is an AI-powered investigative tool featuring 69 prewritten natural language prompts that officers can select, edit, or submit, alongside the ability to input custom prompts.
- The system provides access to 45 distinct tools that query plate scans, camera metadata, arrest records, case files, 911 dispatch logs, ballistics results, and commercial identity databases containing SSNs, dates of birth, phone numbers, email addresses, and associate information.
- Associate tracking uses a co-occurrence algorithm: plates appearing at the same cameras within a default two-minute window are ranked, with a default confidence threshold of 0.75, returning up to 20 associated vehicles from a single target plate.
- Pattern-matching prompts (19 of 69) and plate/name-free searches (14 of 69) enable officers to query by geography, time, and behavior alone—such as vehicles visiting three or more retail locations in three days, multiple banks in a week, or making repeat roundtrips over 14 days.
- A filtering mechanism excludes buses, semi-trucks, work vans, and trailers from pattern searches, potentially creating bias against ordinary drivers while excluding commercial vehicles.
- The justification form for searches defaults to requiring a typed reason but imposes no substantive requirements on content or length, and case numbers need only contain three characters.
- WIRED analyzed the software by examining code served from Flock's login portals, reconstructing the authenticated UI without actual system access; the code contains no instructions given to the AI model itself, leaving server-side behavior undisclosed.
Industry Insight
- The OS Investigate case demonstrates the rapid capability creep possible in commercial surveillance AI, where tools marketed for specific, narrow purposes can be expanded into broad behavioral monitoring systems—companies and their clients should establish clear ethical boundaries and audit mechanisms before deployment.
- The use of preprogrammed natural language prompts lowers the barrier to entry for law enforcement officers who lack technical expertise, but it also encodes the vendor's assumptions about what constitutes suspicious behavior, raising concerns about algorithmic bias and the automation of suspicion without human oversight.
- The commercial data brokerage ecosystem—providing SSNs, relatives, phone numbers, and online accounts—creates a dangerous feedback loop where warrantless surveillance is augmented by private sector data, suggesting that regulators and AI practitioners must address not only surveillance technology but the underlying data supply chains that make it possible.
Disclaimer: The above content is generated by AI and is for reference only.