Last Tuesday’s Quiet Approval
If you weren’t at city council last week, you missed something. Buried in the consent agenda—that collection of items they breeze through in five minutes while everyone’s checking their phones—was approval for a $847,000 AI-assisted surveillance contract. No advance notice in the community newsletter. No separate hearing. Just a line item that appeared, got approved, and moved us into a category most people don’t even know exists yet.
This isn’t unusual anymore. Across the country, law enforcement agencies are adopting AI surveillance tools at a pace that vastly outstrips public awareness or deliberate community choice. According to the ACLU Atlas of Surveillance, over 1,900 law enforcement agencies were already using some form of AI-assisted surveillance by early 2026—everything from automated license plate readers to facial recognition to software that predicts where crimes might happen. The vendor contracts keep expanding. One major company alone, Axon, has contracts with over 17,000 law enforcement agencies globally, generating over $2 billion annually from AI-integrated body cameras and real-time transcription systems.
The speed of this rollout matters. Decisions are being made without the kind of public deliberation that should accompany technology this powerful.
Why This Deserves Your Attention (The Concrete Version)
This isn’t abstract. These systems make mistakes, and the mistakes aren’t random. Research from Georgetown Law’s Privacy and Technology Center found that facial recognition systems misidentified Black men at error rates between 10 and 100 times higher than white men, depending on which system was being tested. In 2024 and 2025 alone, at least seven documented wrongful arrests were traced directly to facial recognition errors. Real people. Real consequences. And the technology is already deployed in communities where you might live.
The pattern matters too. Predictive policing software—another AI tool spreading quietly through law enforcement—concentrates police presence in neighborhoods based on historical crime data, which itself reflects decades of unequal policing. The result is a feedback loop: biased data trains biased algorithms, which guide biased deployment, which creates new biased data. It looks objective. It’s algorithmic. It’s data-driven. And it can systematize inequality in ways that are genuinely hard to challenge because they’re wrapped in the language of neutral mathematics.
That’s not speculation. That’s how these systems have worked in every city where they’ve been deployed and studied. The question for us is whether we want that here, and whether we’re going to make that choice intentionally or let it happen while we’re busy.
What Actually Needs to Happen (And Why It Hasn’t)
The National Institute of Standards and Technology released updated guidance in 2025 specifically for cities buying AI surveillance technology. Their recommendation is straightforward: before deploying any AI surveillance system, conduct a mandatory algorithmic impact assessment. Have experts—and community members—examine what the technology actually does, what it gets wrong, and who bears the cost of those errors. Have the city council debate it openly. Vote on it deliberately. Make the decision auditable.
Sounds reasonable, right? It is. And it’s also rare. According to the Electronic Frontier Foundation’s most recent local policy scorecard, only 11 U.S. cities have actually passed comprehensive AI use policies that require public notice, council approval, and annual audits for surveillance contracts. Eleven. Out of thousands.
Why? Partly because most people don’t know this is happening. Partly because the companies selling these systems are very good at it—they have sales teams, pilot programs, case studies, and relationships with police departments. The people arguing against them are mostly working voluntarily and learning as they go. Partly because city councils genuinely don’t have the expertise to evaluate these systems in-house, so they default to what law enforcement says they need.
And partly because it requires the kind of sustained attention and follow-through that feels impossible when you already have a job and a family and your own problems.
What Changed When Someone Actually Paid Attention
Three months ago, someone in our town—a retired teacher named Marie who, like me, reads the city council minutes because she finds them oddly interesting—noticed a footnote about a police department request for “predictive analytics software.” She did what any reasonable person would do: she asked what that meant. Nobody really had a clear answer. So she asked again. Then she emailed the city council asking them to postpone the vote pending a public hearing and an independent review of the system’s accuracy and bias risk.
The email went to about fifteen people she knew. Three of them forwarded it. The city received maybe sixty emails about something they’d probably expected to sail through. Not a massive movement. Just enough attention to make continuing in silence awkward.
The council voted to postpone. They commissioned an outside review. The review found that the specific software they were looking at had documented bias issues and had contributed to wrongful arrests in at least two other jurisdictions. Based on that report, they’re now designing a policy that requires independent assessment of any surveillance technology before purchase, mandatory public comment, and annual audits. They haven’t rejected the technology entirely—they might still use it—but they’re making that choice deliberately and with their eyes open.
Here’s what changed: someone read the minutes. Someone asked a question. Someone invited their friends to care. And suddenly the decision was no longer invisible.
How to Start, This Week
You don’t need to be a technology expert. You don’t need a lot of free time. You need to do what Marie did, which is to become briefly, intensely curious about one specific thing.
Start here: get on your city’s website and look for the last council meeting agenda. Search for words like “surveillance,” “technology,” “Axon,” “AI,” “predictive,” “recognition.” If you find something, read what they say about it. Email three council members asking them to explain what it does and whether there’s been a bias or accuracy assessment. You don’t need to make an argument. Just ask questions. Questions create accountability because they make silence harder.
If nothing turns up, check the police department’s next budget proposal or their technology roadmap if they have one public. Join your city’s advisory boards or attend one council meeting a quarter. Most of the important stuff that gets approved without debate does so because nobody’s paying attention. Attention is remarkably powerful.
Tell someone else about it. Not to convince them of anything—just tell them what you learned. The spread of real information through small networks is how these policies actually change. Marie didn’t organize a movement. She just talked to friends who talked to friends.
What are you curious about in your city right now? What would change if you read one more thing or asked one more question?