The Contract Nobody Really Read
Your city council probably approved the surveillance camera system in about fifteen minutes. There was a vendor presentation with impressive demo footage, a staff recommendation, maybe one question from a council member about cost, and then it passed. The contract itself was probably hundreds of pages, filed away in a municipal database that requires you to know exactly what to search for to find it. This is not a coincidence. It is, in fact, how most of these deals happen across America.
Over 90 U.S. cities have signed contracts with major vendors like Motorola Solutions’ Avigilon or Axon’s camera analytics platforms that include AI-based behavioral detection features, according to recent tracking work. These systems promise smarter security through artificial intelligence. They promise to detect suspicious activity automatically, alert officers in real time, and make your neighborhood safer. What they also do is something much less visible: they create permanent pathways for data to flow in directions you probably didn’t know about, and they embed algorithmic decision-making into your community’s physical infrastructure.
This article is a map through that contract language. Not because it is reassuring, but because understanding what your city actually signed up for is the only way to do anything about it.
The Data Sharing Problem That Hides in Plain Language
Here is what matters most in these contracts, and what is almost never discussed in city council chambers. According to a 2025 ACLU analysis of 40 municipal AI surveillance contracts, 78 percent included perpetual data-sharing clauses with federal law enforcement agencies. Perpetual means forever. Forever means that if the FBI wants access to footage collected by your neighborhood camera, they can get it without the city council approving that relationship separately. It also means that if the contract gets renewed or extended, this data-sharing relationship continues unless the city actively negotiates to remove it.
This matters because federal agencies do not operate under the same public records laws that municipalities do. A request for footage from your local police department can be submitted through open records laws. A request from the FBI typically cannot. The footage goes into a system, gets connected to facial recognition databases, gets cross-referenced with other federal surveillance tools, and you have no mechanism to know it happened or to object to it.
When you read your contract, look for language about “authorized law enforcement partners” or “federal agency access” or “interagency data sharing protocols.” If you see those phrases and they are not explicitly limited to specific situations requiring separate approval, then you are looking at a perpetual data-sharing arrangement. Write that down. Take a photo of that section. This is the paragraph that matters most.
When AI Gets It Wrong, Who Actually Pays?
In 2024, the city of Shreveport, Louisiana canceled its contract with SoundThinking, a gunshot detection platform. The decision came after a MacArthur Justice Center study found that the system generated false positive alerts 90 percent of the time in predominantly Black neighborhoods. Ninety percent. That means officers responded to alerts that were not gunshots, wasted police resources, created unnecessary police presence in specific communities, and did so on the basis of faulty AI.
This is not theoretical risk. This happens. And when you read your city’s contract, look for what it says about false positives, about accuracy standards, about what happens when the AI gets it wrong. Most contracts do not commit vendors to specific accuracy thresholds. Most do not specify what happens if the system performs differently in different neighborhoods. Most do not require the city to study whether the system is actually making things safer or just making some areas more surveilled.
This is a question you can ask your city council in writing. It is not aggressive. It is reasonable. Ask what accuracy standards the contract requires. Ask what performance data the vendor must provide. Ask how the city will know if the system is working or harming your neighborhood. If the answer is vague or incomplete, then you know something important about what your city actually committed to.
The Accountability Ordinance Your City Probably Doesn’t Have
San Diego passed a Smart Streetlights Ordinance in 2025 that is now considered a national model for surveillance technology oversight. Here is what it requires. Before any new surveillance technology is deployed, there must be a public 90-day review period. During that time, community members can read about the technology, ask questions, and provide feedback. After that review period, the city council must hold a vote, and the vote must be a supermajority rather than a simple majority. It creates friction in the process. It creates time for people to pay attention. It makes it harder to sneak these systems into neighborhoods without anyone noticing.
According to research from the Georgetown Center on Privacy and Technology, fewer than 15 percent of U.S. cities with AI surveillance contracts had enacted any algorithmic accountability ordinance at all. Your city is probably in that 85 percent. This does not mean it is impossible to change. It means there is an opportunity.
Start by checking whether your city has any surveillance technology oversight ordinance. Check your city council website. Check your municipal code. If it does not exist, then you have something concrete to push for. You can point to San Diego. You can argue that surveillance technology should be subject to the same approval process that any major city infrastructure gets. You can make this argument to city council members, and you can make it in writing for the public record.
What You Can Actually Do This Week
First, find your city’s surveillance contract. Call your city clerk’s office. Ask specifically for the AI camera system contract and the vendor agreement. Ask for all amendments and extensions. These are public records. You have a legal right to them. You might need to file a formal public records request, but start with a phone call.
Second, read through the contract and look specifically for these things: data sharing clauses, accuracy standards, community notification requirements, and any language about algorithmic audits or performance monitoring. You do not need to understand every word. You are looking for presence or absence. You are looking for what commitments are actually written down and what is left to vendor discretion.
Third, take what you find and bring it to your next city council meeting. Attend in person if you can. Speak during public comment. Ask specific questions about the contract language. Ask what accuracy standards the vendor must meet. Ask whether your city conducted any community input before signing. Ask what happens if you discover the system is performing differently in different neighborhoods. These questions matter because they go into the public record, and they make future council members aware that people are paying attention.
Fourth, connect with any local community groups already working on surveillance or policing issues. This is not a problem you need to solve alone. Neighborhood associations, civil rights organizations, community technology groups are already engaged with this work. Bring them your contract questions. Bring them your concerns. Build something together.
The contract your city signed is not destiny. It is a starting point. It reflects choices that your city made, and future city councils can renegotiate, amend, or cancel it. But only if enough people know what is actually in it and decide it matters enough to do something about.
If you read your contract and have questions about what the language means, bring those questions to your next neighborhood association meeting. Post them to your local community Facebook group. Tag your city council members. Start the conversation. The point is not to have all the answers right now. The point is to know what questions to ask and to ask them loudly enough that someone has to answer.
