Flock Safety's AI searches can start with behavior, not a suspect

Why it matters

Flock Safety is moving from recording vehicle evidence to generating investigative targets, forcing Langley to prove that product safeguards can keep pace with AI-powered police search.

Garrett Langley (@glangley) has built Flock Safety into a national police technology platform by giving officers vehicle evidence. WIRED reported on August 19 that Langley's next product can begin an investigation without a known license plate, person, or crime.

The AI system, originally called Nightshift and now known as OS Investigate, lets officers search for people and vehicles through patterns of movement. According to WIRED, the system can connect vehicle scans with case files, arrest records, 911 dispatch logs, ballistics results, and commercial identity databases.

The expansion takes Langley a long way from Flock Safety's first prototype, an Android phone placed in a waterproof box after burglaries hit his Atlanta neighborhood. Langley, a Georgia Tech electrical engineering graduate, founded Flock Safety with Matt Feury and Paige Todd in 2017 and joined Y Combinator's Summer 2017 batch. The original pitch was direct: capture the plate information police said they needed to investigate property crime.

OS Investigate turns that evidence system into a pattern engine. For investigators, the intended payoff is speed. Searches that once required moving among camera records, dispatch software, case files, and identity databases can be assembled through a chat interface. The risk comes from the same feature: officers can ask the software to choose who deserves closer scrutiny.

From vehicle evidence to behavioral search

Some OS Investigate prompts start with familiar police work. An officer can enter a plate and ask for associated vehicles seen near it.

Other prompts reverse the search. One asks for possible witnesses by finding vehicles most frequently seen in a neighborhood over the previous 14 days. Another looks for vehicles visiting three or more retail locations in three days. Similar searches target repeated visits to banks, overnight stops at gas stations, or round trips between cities.

WIRED reported that 19 prompts focus on patterns, while 14 require no plate, name, or physical description. An officer supplies a place, a period, and a behavior, and OS Investigate returns people or vehicles that fit.

The system can then enrich a plate with names, addresses, relatives, phone numbers, online accounts, and prior appearances in police records. A separate prompt begins with people arrested more than twice in two years, maps their homes, retrieves service calls at those addresses, and orders a detailed "workup" on three people selected by the workflow.

Those outputs are investigative leads. An inferred association between two vehicles does not establish that their owners know each other, and repeated trips through a commercial district do not establish criminal conduct. The convenience of a prewritten prompt can make an uncertain statistical relationship look like a routine database result.

WIRED reconstructed portions of the interface from files exposed through Flock Safety's login pages. The code shows prompts, available tools, search forms, and intended workflows. It does not reveal the model's underlying instructions, the server-side checks applied to requests, which tools are active for current testers, or how the system responds in operation.

Flock Safety told WIRED that OS Investigate remains in development with a small group of law enforcement partners. Spokesperson Paris Lewbel described it as separate from Flock Safety's license plate reader product and said the observed capabilities may change before a wider release.

Langley's next platform bet

The product fits the direction Langley has pursued for years. Flock Safety has expanded from fixed plate readers into video cameras, gunshot detection, drones, 911 software, and tools for real-time crime centers. OS Investigate supplies the layer that can query those records together.

Investors have financed that move. Flock Safety's March 2025 financing raised $275 million at a $7.5 billion valuation, led by Andreessen Horowitz, with backing from Greenoaks Capital and Bedrock Capital. Meritech Capital, Matrix Partners, Sands Capital, Founders Fund, Kleiner Perkins, Tiger Global, and Y Combinator also participated.

The financing announcement said the investment would fund product innovation, research and development, and U.S.-based manufacturing, with investigative intelligence among the areas targeted for development. OS Investigate shows the strategic destination: Flock Safety wants to own the software where officers decide what to investigate, extending its business beyond camera feeds.

The market is heading in the same direction. Axon introduced fixed plate-reading cameras and additional AI tools in 2025, tying vehicle recognition to its Fusus real-time operations platform. Police vendors are competing to combine dispatch, evidence, video, and records inside a single searchable system. Flock Safety enters that contest with a camera network that WIRED places in more than 6,000 communities.

The guardrails have to move as fast as the product

Langley's central product challenge is governance. Flock Safety has long emphasized that its plate readers focus on vehicles and do not use facial recognition. OS Investigate adds identity records, physical-description searches, movement analysis, and software-generated associations around that vehicle data. The distinction between tracking a plate and tracking a person becomes difficult to preserve once the same interface connects one to the other.

Flock Safety is tightening controls after documented misuse. The Washington Post documented more than 40 officer misuse cases, including alleged stalking through the network.

On August 13, Langley announced new safeguards, including a recommended seven-day default retention period, mandatory audit assistance for law enforcement customers, automatic lockouts for abnormal activity, and required case codes by the end of 2026. Existing customers can retain locally approved data-retention periods, and emergency searches can bypass the case-code requirement subject to review.

Those controls were designed around searches initiated by people. OS Investigate adds automated lead generation, editable natural-language prompts, and chains of searches across several databases. Effective oversight will have to capture the officer's query, each tool the AI invoked, the records returned, the inferences generated, and the actions taken afterward.

Local resistance is already shaping Flock Safety's sales environment. CBS News reported on February 24 that Mountain View voted to terminate its Flock Safety contract, Los Altos Hills was among Northern California cities that had terminated automated license plate reader contracts, and Santa Clara County supervisors discussed changing county policies governing the cameras. In Texas, the Bandera City Council voted 3-2 on May 12, 2026, to terminate its Flock Safety contract.

OS Investigate gives Langley a technically credible answer to a real police problem: investigators have too much disconnected data and too little time. It also gives officers a system capable of producing subjects from ordinary movement. Langley's next proof point is whether Flock Safety can place enforceable limits inside the product before an AI-generated lead reaches a person who was never suspected of a crime.