AI News
23 Aug 2026
Read 11 min
How Flock OS Investigate works and what it reveals
How Flock OS Investigate works to map driver networks and expose privacy risks for local communities.
How Flock OS Investigate works: It links license-plate camera hits, police records, and commercial databases into a chat-style AI that runs preset and custom searches. Officers can map movements, spot “associates,” and pull personal details, even without a plate or name, raising major questions about privacy, accuracy, and oversight.
Flock Safety built its business on automated license-plate readers (ALPRs). Now its test-stage AI, OS Investigate (formerly Nightshift), goes further. Based on code WIRED examined from Flock’s own site, the tool lets police search across plate scans, case files, dispatch logs, and identity data with simple prompts. It can suggest “witnesses,” show who travels together, and surface names and addresses tied to plates. The company says it is still refining features, but the direction is clear.
License-plate scans and camera metadata from a network used in 6,000+ communities
Arrest records, case files, and 911/dispatch logs
Ballistics results
Commercial identity files with names, addresses, relatives, phone numbers, emails, dates of birth, and Social Security numbers
Instead of only checking a known plate against a hotlist, the system can start with a place, a time window, and a behavior pattern, then return vehicles and linked people that match those conditions.
“Find me witnesses” by listing vehicles most often seen in a neighborhood over specific days and times (excluding whitelisted vehicles)
Show “associates” of a car by finding plates that appear at the same cameras nearby within minutes, multiple times
Identify vehicles that visited several retail stores in three days, multiple banks in a week, or gas stations between midnight and 5 am
Trace cars that traveled city-to-city and back within a week or made repeated round trips over 14 days
Search by physical description alone (for example: “male, ~6ft, black hoodie, forearm tattoo”) near a drawn area on a map
The tool can filter out buses, semis, work vans, and trailers to avoid flagging obvious commercial routes, yet still include ordinary drivers who make similar stops.
Registered vehicles and prior mentions as a suspect
Home addresses, relatives, phone numbers, and online accounts
In practice, this means the AI can move from a movement pattern to a plate, then to a name and address, often without starting from a known suspect or clear crime.
Help solve vehicle thefts and robberies faster
Spot repeat routes and potential getaway patterns
Surface possible witnesses quickly after an incident
Connect leads across city and state lines
Starting searches without a plate, name, or defined crime increases dragnet risk
Prompt freedom lets officers steer the AI toward vague “find criminal patterns” hunts
Description-only searches can amplify bias (by race, sex, age, build, tattoos)
Large-scale geofence-style sweeps may implicate people who did nothing wrong
Recent reporting has already documented misuse of ALPR systems by some officers, including stalking and searching for political and personal reasons. While Flock has announced new safeguards like abnormal-search monitoring, auto lockouts, and required case codes, many agencies still control who gets access, what they can search, and whether sharing spans jurisdictions.
Set strict legal thresholds (warrant or written articulable suspicion) for pattern-of-life searches
Ban description-only dragnets and protest- or clinic-related queries
Mandate specific, reviewable justifications and full case numbers on every search
Enable robust audit tools, with automated anomaly alerts and quarterly public reporting
Limit data retention and enforce minimization (no saving non-hit scans beyond a short window)
Constrain cross-agency sharing and require formal agreements with audits
Train users on bias, false positives, and due process standards
Include community oversight boards and publish transparency dashboards
How Flock OS Investigate works
The AI agent and its data connections
At its core, OS Investigate is a chat-style assistant that automates police queries across many datasets. The code WIRED reviewed shows 45 tools behind the scenes and 69 prewritten prompts officers can click, edit, and run. It can also accept freeform instructions. It can reach:Prompt-driven searches and sample workflows
Flock’s preset prompts act like recipes. Officers can accept them as-is or change the wording. Examples include:What the tool can reveal
Pattern-of-life and “associates” logic
The “associates” feature shows how the model builds relationships. According to the code, the AI counts co-appearances of plates at the same camera within a two-minute window. If another car appears with a target three or more times and passes a default “confidence” threshold of 0.75, it is ranked as an associate. A single seed plate can yield up to 20 linked vehicles.From plates to people
When OS Investigate lists plates, other tools can run a one-command “workup” that bundles what the agency and commercial sources hold on a person:How Flock OS Investigate works within police workflow
Case justifications and controls
Before a search, the interface can require a short reason and, if enabled, a case number. WIRED’s review suggests these fields can be minimal (for example, any three characters for a case code), and it is unclear what server-side checks exist. Flock says capabilities and workflows may change before wider release.Speed and scope
The company’s CEO has described demos where the AI searched citywide cameras for cars tailing an armored truck and pulled linked arrest records in under a minute. The platform logs around 20 billion plate scans each month and counts about 140,000 monthly active users, showing the potential scale of analysis if the AI is rolled out widely.Benefits, risks, and the legal tightrope
Potential gains for cases
Supporters say the system could:Privacy, bias, and “fishing expeditions”
Civil liberties experts warn the tool shifts from targeted checks to broad pattern-of-life surveillance. Key concerns include:Governance gaps to close now
Practical steps for cities and agencies
If agencies test or adopt the system, they should harden policy and oversight first:What to watch next
Flock is hiring for deeper “agentic” features, like automated lead generation and cross-camera correlations. As these systems grow, the balance between fast investigations and constitutional rights will hinge on transparent rules, real audits, and firm legal constraints—before the tech normalizes dragnet surveillance. In short, understanding how Flock OS Investigate works is essential for anyone weighing its promise against its risks. It can turn movements into identities and networks in minutes. Without clear limits, that power could shift policing from targeted leads to open-ended surveillance.(Source: https://www.wired.com/story/flock-safety-os-investigate/)
For more news: Click Here
FAQ
Q: What is OS Investigate and what can it do?
A: How Flock OS Investigate works: it links license-plate camera hits, police records, and commercial identity databases into a chat-style AI that runs preset and custom searches. It can map vehicle movements, surface “associates,” and convert plates into names, home addresses, and related records even without a starting plate or name.
Q: What sources of data can OS Investigate access?
A: The code WIRED reviewed shows OS Investigate can reach license-plate scans and camera metadata from a network used in more than 6,000 communities, arrest records, case files, 911/dispatch logs, ballistics results, and commercial identity files containing names, addresses, phone numbers, emails, dates of birth and Social Security numbers. Those connections are described among the system’s roughly 45 backend tools and 69 prewritten prompts.
Q: How do the preset prompts and freeform searches work in OS Investigate?
A: Flock ships OS Investigate with 69 prewritten prompts that officers can click, review, edit, and submit, and the interface also accepts freeform instructions typed into a chat box. The prompts act like recipes for searches such as finding vehicles most seen in a neighborhood during specific days and times, listing a car’s top “associates,” or tracing vehicles that visited multiple locations.
Q: Can investigators start searches without a plate or a suspect’s name?
A: Yes; instead of beginning with a known plate or person, an officer can supply a location, a time window, and a pattern of behavior and the system is designed to return matching vehicles and linked people. Flock says the product is still in development, but WIRED’s examination of code shows those workflows are preloaded in the interface.
Q: How does the system determine “associates” of a vehicle?
A: The code describes that the system counts other plates that appear at the same camera within a two-minute window and reports any car that co-appears three or more times and exceeds a default confidence threshold of 0.75, returning up to 20 linked vehicles. That lets a user supply a single seed plate and receive a ranked list of vehicles considered associated by movement patterns.
Q: What is a “workup” and what information does it provide?
A: A “workup” is a one-command background check that starts with a name and date of birth and returns what an agency’s records and commercial data hold, including registered vehicles and prior listings as a suspect. On a secondary screen it can surface home addresses, relatives, phone numbers, and online accounts as described in the code WIRED reviewed.
Q: What are the main privacy and legal concerns about OS Investigate?
A: Experts say the tool shifts policing from targeted hotlist checks to broad pattern-of-life surveillance, enabling geofence-style dragnets, description-only searches, and queries without a defined crime that risk implicating innocent people and amplify bias. The reporting also notes the Supreme Court has held people retain a privacy interest in location records and that the Fourth Amendment can apply when police demand those records from companies.
Q: What safeguards has Flock announced and what governance steps do experts recommend?
A: Flock announced it will monitor for abnormal searches, automatically lock out flagged users, and require officers to attach case codes by the end of the year. Experts and the article recommend agencies require warrants or articulable suspicion for pattern searches, ban description-only dragnets, mandate full justifications and audits, limit data retention and cross-agency sharing, and include training and community oversight before broader deployment.
Contents
