Why public safety technology programs get cancelled, and what the agencies that keep them do differently
Public safety technology programs rarely fail because the technology does not work. They fail when the agency cannot show its work: no published policy, no audit trail, a contract structured to avoid council review, or a vendor data-sharing arrangement the department did not know it had. The agencies that survive a hostile council meeting are the ones that disclosed early, published the policy, audited their own use, and could answer specific questions with specific records.
Someone breaks a car window in a shopping center lot and takes a laptop off the seat. No witness, no suspect description, nothing worth printing. Twenty years ago that report got taken, filed, and closed. Everybody knew it. The victim knew it.
Today that case sometimes closes, and when it does, it usually closes for one reason: something recorded the plate leaving the lot. A camera, a reader, a neighbor's doorbell. The follow-up is real police work, but the thread that made it possible came from a device on a pole.
That is the part worth keeping in view while the rest of this plays out. The argument over public safety technology is usually framed as privacy against surveillance. On the ground it is more often a question about which cases get solved and which victims get a phone call that goes nowhere.
What is actually happening
Cities are pulling this equipment. Not everywhere, and not most places, but enough that it is no longer a curiosity.
NPR counted at least 30 localities that deactivated cameras or cancelled license plate reader contracts between the start of 2025 and February 2026. Austin let its contract lapse. Evanston, Oak Park, Denver, Sedona, Coralville, and San Marcos ended theirs. In July 2026 the Los Angeles Police Department announced it would not renew, citing civil rights concerns. Chicago decommissioned its gunshot detection system in September 2024 after a city inspector general found that 9.1 percent of alerts produced documented evidence of a gun-related offense.
The trigger for the license plate reader wave is well documented and worth understanding precisely, because it was not a general objection to the technology. In May 2025, reporting based on public records showed officers running plate searches with reasons typed into the field like "immigration" and "ICE." Days later came a search logged as "had an abortion, search for female," run across tens of thousands of cameras in states where that care is legally protected. In August 2025, the Illinois Secretary of State audited the vendor and found it had given a federal agency access in violation of state law, through a pilot the company's own leadership did not know was running.
Read the council transcripts from the cancellations that followed and the objection is remarkably consistent. It is rarely "we do not want plates read." It is "we do not know who else is reading them."
The pattern in the programs that died
Four cases, four technologies, one thread.
| Agency | What happened | The actual cause |
|---|---|---|
| New Orleans | Predictive policing partnership ended in 2018 after a news investigation | Structured as a philanthropic donation, so it never went through procurement. Council members and city attorneys did not know it existed. |
| Nashville | Camera integration platform rejected by council in December 2024 | The original contract was written at $175,000, just under the $250,000 threshold that would have required council approval. When the department later needed council, that structure became the story. |
| Oakland | Privacy commission moved in 2021 to terminate plate readers entirely | No annual reports and no audits for five consecutive years, and the commission found the department had represented that audits were being performed when its system could not perform them. |
| Baltimore | Aerial surveillance ended in 2021, later held unconstitutional | Ran without a warrant framework. A federal appeals court held that accessing the data was a search. |
None of those programs died because the equipment failed. Three of the four died because of how the agency got the equipment in the door.
What the agencies that kept it did
This is the more useful list, and it gets less attention.
- Cleveland extended its plate reader contract 9 to 6 in July 2026, and the amendments are the whole story. Council barred the Northeast Ohio Regional Fusion Center from reaching city data, required a warrant for any outside agency access, shut off the vendor's national lookup, and ordered a public transparency portal plus quarterly reports to council. The extension runs six months and the city is taking bids from other vendors in the meantime. The capability survived. The vendor relationship may not.
- Rockford, Illinois approved 27 more readers and 7 more cameras on a 13 to 1 vote in March 2026, a four year contract worth about $444,000. The department's case was numbers. Its homicide clearance rate was roughly 47 percent in 2022, before the equipment, and 90 percent or better in each of the three years since. The city also reported that 75 percent of its 2025 homicide arrests relied on plate reader data.
- Lexington, Kentucky kept its camera platform by writing the policy with the ACLU, the NAACP, and the local human rights commission in the room. The resulting policy limits authorized users, keeps data inside the crime center, and allows officers to pull private camera feeds only when they are already responding to a call at that location.
- Seattle runs every surveillance technology through a published impact report, a public comment period, and a council vote. When a detective was found to have acquired facial recognition software on his own, outside the process, the result was a reprimand and a rule change. In a department without that framework, the same facts produce a lawsuit.
Notice what the survivors have in common. None of them won by defending the status quo. They won by showing up with limits already written.
One honest caveat, because the primary sources will not let this be tidier than it is. Several of those renewals passed over heavy public opposition. In Berkeley, public comment ran roughly 100 against to 6 in favor and the renewal still passed 5 to 4. In Huntington, West Virginia, more than 50 speakers appeared and all but one opposed. The guardrails did not make these programs popular. They made them survivable, and they gave councils something to vote for other than the vendor's assurances.
The national picture, taken together, is not a wave of cancellations. It is closer to this: agencies are cancelling at the margins, constraining in the middle, and expanding at the state level, where there is no city council to lose a vote in.
The part that complicates this
Process is not a guarantee, and any article that tells you otherwise is selling something.
Baltimore's aerial program had a formal pilot, an independent civil rights audit, and a RAND evaluation that found a statistically significant clearance improvement on harder cases. It went through an open process and it still ended, and a federal appeals court still found it unconstitutional.
The honest version is narrower than "be transparent and you will be fine." Process does not determine whether your program survives. It determines whether you can defend it when it is challenged, and whether a mistake ends as a correction or as a ban. Seattle's unauthorized software use ended in a reprimand. Baltimore's ended in a published opinion.
What the public actually thinks, as opposed to what the loudest week suggests
Support for this equipment is high and it is mission-specific. Across two decades of polling the pattern does not move. Search and rescue polls in the eighties and nineties. Locating a missing child polls near the top. Every instrument implying routine or continuous monitoring drops sharply, in some polls to a fifth. That gap is a structural feature of public opinion, not a communications failure you can fix with a better brochure.
Two findings are worth sitting with.
The first is about awareness. Chula Vista is the program the Justice Department holds up as the model for community engagement, and it has published a public flight log since 2021. A 2024 survey found that only 34 percent of residents knew the program existed, and 16 percent knew any policy governed it. A portal is something you can point at when challenged. On its own it does not appear to change many minds, and anyone who tells you it will is guessing.
The second is about the vandalism you may have seen in the news. Cameras have been cut down, shot, and painted over in at least twenty states. It is real, and departments are charging it as a felony. It is also, against roughly 120,000 cameras deployed nationally, somewhere in the range of 100 to 150 devices. Almost every case is one person or two. The organized advocacy groups in this space run town halls and public records requests, not saw crews.
What is genuinely notable is not the count. It is who is involved. A constitutional conservative outside Houston, a progressive in Massachusetts, and a rural independent in South Carolina have arrived at the same position. That crossover is the part command staff should track, because an agency can plan around opposition from a known quarter. Losing your own base is a different problem.
What the evidence actually supports
Be careful here, because the marketing in this space has gotten bad enough that it is now a liability for the agencies buying it.
The Justice Department's own research clearinghouse rates license plate recognition Ineffective as a crime deterrent, finding no statistically significant effect on general crime, auto-related crime, or auto theft. The most current academic review, published in late 2025, went through 200 real case files where plate readers were used and found that in nearly three quarters of them the reader did not contribute in any meaningful way to resolving the investigation. About 14 percent of the reader data in that sample was a false positive.
Vendors have also been caught inflating. In San Marino, California, the company advertised an 80 percent drop in residential burglaries. Burglaries actually went up 5 percent. The police chief's response when a reporter showed him the real numbers was that he needed to talk to their marketing people.
Carry a vendor's statistic into a council meeting and let someone else check it, and you lose the room and the program in the same evening. Bring your own numbers.
Here is what the evidence does support, and it is narrower and more useful.
- Berkeley PD reported 37 stolen vehicle recoveries, 58 arrests, and 121 cases supported by plate readers in 2025.
- The manhunt for the man who killed an MIT professor and opened fire at Brown University tracked his rental across three states, after he had already swapped the plates.
- The National Center for Missing and Exploited Children credits the technology with contributing to the safe return of more than 1,000 missing people.
- Gunshot detection consistently shows faster response and faster medical transport, and consistently fails to show fewer shootings or better clearance.
- Real time crime centers show improved case clearance in the published research, and nothing on crime reduction.
- Drone first responder programs show real response time improvements and a meaningful number of calls that never needed a patrol unit at all.
None of that is crime prevention. It is case resolution, in a narrow band of cases: the stolen vehicle, the manhunt, the missing person, and the property crime with no witness and nothing worth lifting.
That last one is the category worth defending out loud. Somebody breaks a window in a parking lot and takes a bag off the seat. There is no suspect description and no usable print. In that case the plate read is not one lead among several. It is the only one that exists. Take it away and the case does not get worked harder. It does not get worked at all, and the person it happened to learns that is simply how it goes now.
An agency that loses this equipment does not return to 2005 policing with 2005 expectations. It returns to 2005 capability with 2026 expectations, and that gap lands on the victims of exactly the crimes that leave the least behind.
What to do with this if you run a program
- Disclose before you deploy, not after. Every failure above traces to a program that was discovered rather than disclosed. In California, agencies with drones are already required to publish a policy, hold a hearing, and report annually under AB 481. Twenty six jurisdictions nationally have oversight ordinances that go further.
- Audit your own use, and be able to prove it. A 2020 state audit of four California agencies found that not one had audited its own plate reader searches. Misuse in this field is overwhelmingly discovered by victims rather than by internal review, which is the worst possible way for it to surface.
- Know exactly who else touches your data. The cancellations of 2025 and 2026 were driven by data sharing agencies did not know they had agreed to. Read the vendor's default sharing settings, and read them again after every product update.
- Do not structure a purchase to avoid review. It works once. It becomes the headline the second time.
- Bring the critics in early. Lexington wrote its policy with the ACLU at the table and kept its system. That is not capitulation. It is the cheapest form of pre-litigation review available.
- Come to the vote with limits already written. Every agency that kept its technology in 2026 did so by proposing constraints, not by resisting them.
The bottom line
The technology is not the variable. Two departments can run identical equipment and end in opposite places, and what separates them is almost never the hardware. It is whether the agency can sit in front of a council, answer a specific question with a specific record, and show that it was already holding itself to the standard someone else is now demanding.
That is not a public relations exercise. It is the difference between keeping the thing that closes the smash and grab, and explaining to the next victim why it does not.
Sources
Key claims above link to their source inline. The primary documents worth reading in full:
- National Institute of Justice, CrimeSolutions — federal rating of license plate recognition as a crime deterrent.
- Los Angeles Police Commission, Office of the Inspector General — plate reader review, July 2026, including alert outcomes and stop tactics.
- California State Auditor, Report 2019-118 — audit of plate reader practices at four California agencies.
- Chicago Office of Inspector General — review of gunshot detection alert outcomes.
- Oakland Privacy Advisory Commission — findings and motion on plate reader reporting and audits.
- Metro Nashville Community Oversight Board — report on the camera platform contract and the absence of a governing policy.
- Leaders of a Beautiful Struggle v. Baltimore Police Department — Fourth Circuit, en banc, on aerial surveillance and the Fourth Amendment.
- City of Seattle surveillance ordinance — the impact report and public comment process, with current reviews posted.
- ACLU, Community Control Over Police Surveillance — oversight ordinance adoption and the model bill.
Figures reflect reporting and published research available at the time of writing. Where an agency reported its own results, that is noted in the text.
Almost always over data sharing rather than the cameras themselves. Reporting in 2025 showed officers running plate searches with immigration-related reasons typed into the search field, and one search logged in connection with an abortion investigation run across tens of thousands of cameras in states where that care is protected. An Illinois state audit then found the vendor had given a federal agency access in violation of state law through a pilot its own leadership did not know about. The council objection in most of these votes is not that plates are being read. It is that the agency could not say who else was reading them.
It helps resolve cases more than it prevents crime, and the distinction matters. The Justice Department's own research clearinghouse rates license plate recognition Ineffective as a crime deterrent, and a 2025 academic review of 200 case files found the readers did not meaningfully contribute in nearly three quarters of the cases where they were used. Real time crime centers show improved case clearance and no measurable effect on crime rates. Gunshot detection shows faster response and faster medical transport, and no reduction in shootings. Where this equipment earns its place is narrower: the stolen vehicle, the manhunt, the missing person, and the property crime with no witness and nothing to lift, where a plate read is often the only investigative thread that exists.
Community Control Over Police Surveillance ordinances require an agency to get governing body approval before acquiring or using surveillance technology, publish an impact report and a use policy, and report annually. Twenty six jurisdictions had adopted one as of late 2024. In California, AB 481 imposes a related set of duties on agencies with drones and other listed equipment, including a published policy, a public hearing, an annual report, and a community meeting.
The agencies that kept their systems in 2026 did it by proposing limits rather than defending the status quo. Cleveland's renewal passed because it added a warrant requirement for third parties and limited fusion center access. Lexington kept its platform by writing the policy with civil liberties groups in the room. Coming to the vote with constraints already drafted, real usage numbers, and an audit you can produce is a materially different position than arguing the technology is fine.
There is no good evidence that it does on its own. Chula Vista has published a public drone flight log since 2021 and a 2024 survey still found only 34 percent of residents were aware the program existed and 16 percent knew a policy governed it. A portal is valuable because it gives you something specific to point at when you are challenged, and because building one forces the internal discipline of tracking what you are actually doing. Treat it as a defensibility measure rather than a persuasion campaign.
If you are building the policy and the reporting side of a program before you take it to council, we can show you what the transparency piece looks like in practice.
Sukh Bhela is a California police sergeant who has served as a UAS operator, UAS supervisor, and incident commander during critical incidents. His experience leading patrol operations and integrating drone technology into public safety responses led him to found BabbarOps, where he builds tools for live situational awareness and incident command. He writes about policing, drone operations, leadership, and the technology shaping the future of emergency response.
The views expressed here are the author's own, written in his personal capacity. They do not represent, and are not made on behalf of, any law enforcement agency or employer.
This article is general information and not legal advice. Surveillance oversight requirements, public records obligations, and data sharing rules vary by state and by jurisdiction. Figures cited reflect reporting and published research available at the time of writing. Confirm current requirements with your agency counsel.