Boston abandons Flock after nationwide sharing violated its contract

Why it matters

Flock's growth depends on national network effects, while city contracts depend on enforceable local control. Boston shows that permissions and audit systems now matter as much as camera accuracy.

Boston has abandoned Flock Safety after a city report found that Flock Safety enabled nationwide access to police camera data despite a contract requiring sharing to be turned off. Mayor Michelle Wu confirmed the decision before Boston published its annual surveillance report on September 14th, according to reporting from Ars Technica and WBUR.

The loss lands squarely on Garrett Langley, who built Flock around the idea that inexpensive cameras and a shared software network could give police usable evidence faster than conventional security systems. Langley and co-founders Matt Feury and Paige Todd started with an Android phone inside a waterproof box. Boston's finding shows how far the harder engineering problem has moved: the product now includes permissions, audit trails and sharing controls that must work exactly as promised across a national network.

Three days of open access ended a five-month trial

Boston deployed about 45 Flock automated license-plate-recognition cameras under a no-cost pilot that ran for 156 days, from April 1st through September 4th, 2025. The contract restricted access to trained Boston Police Department personnel, while Boston police said they confirmed with Flock Safety that data-sharing settings were off.

Within three days, Boston police learned that other law-enforcement agencies could access license-plate and vehicle information collected by the cameras through Flock's nationwide lookup function. The city surveillance report attributed the exposure to a Flock "vendor error." Boston had the function disabled, after which agencies had to submit written requests for access.

The open-access period was short, but Boston's response reflects the sensitivity of the data involved. A plate read can place a vehicle at a specific location and time while adding visible characteristics such as color, make and model. The value of that record in an investigation is also what makes an incorrectly configured sharing permission consequential.

During the remainder of the pilot, Boston received 77 access requests from agencies including the Drug Enforcement Administration, the Bureau of Alcohol, Tobacco, Firearms and Explosives, Massachusetts State Police and nearby police departments, WBUR reported. Boston police conducted 11,004 searches of their own, attaching a crime-related case number only about half the time. The city report found no complaints or documented misuse during the pilot, and Boston deleted the data and removed the equipment within 30 days of the trial ending.

Those findings give Flock Safety a fair distinction to draw: Boston identified an access-control failure, rather than documented abuse of Boston data. For a public-safety system, the distinction does not repair the underlying trust problem. Boston had negotiated a specific boundary, checked the setting and discovered that the boundary did not hold.

Flock's network became its advantage and its liability

Langley started Flock in Atlanta after break-ins in his neighborhood left police without actionable evidence. Commercial plate readers could cost roughly $25,000 per camera, according to the founders' account. Todd, a former Home Depot employee who had already worked with Langley on two businesses, cold-called 20 neighborhood associations and signed two early customers. Langley and Feury handled the early product, and the founders joined Y Combinator's Summer 2017 batch.

The founders made deployment cheaper, then made each deployment more useful by connecting it to other cameras and agencies. Flock now markets a national license-plate-reader network alongside video cameras, gunshot detection, drones and investigation software. Cross-agency access is therefore embedded in Flock's commercial advantage. It also concentrates governance risk in account settings that can cross city and state lines.

Investors priced that network aggressively. Andreessen Horowitz led a $275 million financing announced on March 13th, 2025, at a reported $7.5 billion valuation. Greenoaks, Bedrock, Meritech Capital, Matrix Partners, Sands Capital, Founders Fund, Kleiner Perkins, Tiger Global and Y Combinator participated. Flock told TechCrunch at the time that annual recurring revenue had passed $300 million, a figure that remains self-reported.

Boston is the clearest possible customer test for that valuation thesis. A national network is commercially powerful only while local governments believe they retain control over their piece of it. RuntimeWire previously reported that more than 2,000 organizations could access plate data shared by Alpharetta, Georgia. Boston had explicitly declined that model and still found its data inside the nationwide lookup system.

Flock's product ambitions have also expanded beyond plates. RuntimeWire reported in August that Flock searches can begin with vehicle behavior rather than a known suspect, and in September that its AI watchlist interface supports searches for people by description. Each added search mode raises the cost of weak access controls because the system can generate investigative leads from broader and less specific inputs.

Langley is turning optional safeguards into requirements

Flock Safety has already begun changing the controls implicated by the Boston report. In an August 13th post, Langley said Flock would recommend seven-day retention instead of 30 days, require case codes for law-enforcement searches by year-end, mandate misuse-detection software and introduce automatic lockouts for abnormal activity. Flock also added offense-based controls intended to let a city determine which categories of outside investigation can reach its cameras. Those are Flock's stated changes, rather than independent validation of how the controls perform.

The timing matters. Boston's configuration failure happened in April 2025, before Flock announced those changes. The report became public as Langley is trying to persuade cities that local policy can be translated into enforceable software permissions. Boston's audit also supplied a direct case for one of his decisions: when case numbers were optional, officers included them in only about half of searches.

Flock still has room to prove that the newer controls match Langley's promise that customers decide who can access their data. Boston offers a concrete standard. A sharing prohibition must survive deployment without depending on police discovering an incorrect setting after cameras begin collecting records.

Boston has not rejected license-plate readers themselves. Police began testing a Motorola Solutions system in February 2026 and Axon hardware in June. Wu said Boston could use the technology under a publicly approved policy governing access, retention, privacy and civil liberties.

That leaves Langley with a product challenge rather than a lost argument over whether plate readers can help police. Boston still wants the capability. It chose to test Flock's rivals because the city's contract and Flock's settings produced different outcomes. For a founder who turned cheap camera hardware into a multibillion-dollar network, closing that gap is now core product work.