The town of Matthews added AI surveillance cameras to a town board meeting agenda without telling anyone. Commissioner Urban moved to add it. The motion passed. The police chief presented for 20 minutes about 25 AI cameras covering downtown and Stumptown Park. The board asked about hackability. Someone laughed.

Then the vendor said the price was going up after July 3rd, so they called a special meeting nine days later on the eve of the Fourth of July holiday weekend. Minimal public notice. One person showed up. One person came to speak about whether their town should build AI surveillance infrastructure. She raised racial bias, civil liberties, and not to let a vendor's deadline make the decision for them. The board thanked her. Then they talked about the California choice of law provision in the contract. This is what power looks like when it has stopped worrying about being seen.

What The Numbers Say

The woman who showed up to that special meeting was not being dramatic. She was citing documented fact.

AI facial recognition systems misidentify Black people at error rates as high as 34%, according to peer-reviewed research, while error rates for white men fall below one percent. As of March 2026, there have been at least nine documented wrongful arrests in the United States tied to AI misidentification, the majority of them involving Black people.

In Hampton Roads, Virginia, a Christopher Newport University study found that Flock automated license plate cameras (the same technology now operating across Mecklenburg County under a county government contract) surveil Black neighborhoods at nearly four times the rate of white neighborhoods. 75% of high-poverty census tracts in that study contained at least one camera. Fewer than 40% of low-poverty tracts did.

Surveillance cameras don't get placed randomly. They get placed where policing has historically been concentrated, which means the data those cameras generate reflects the same racial biases built into every prior policing decision. The algorithm learns from that data. The algorithm is then called objective. This is not a flaw in the system. It is the system.

In Atlanta (which has its own Cop City and its own expanding surveillance network) a 2025 mapping project found approximately 124 surveillance cameras for every 1,000 residents, more than any city in the world outside a handful in China. Atlanta police have already used social media monitoring and networked cameras against Cop City opponents. Civil liberties researchers warn that what gets rehearsed in police training facilities does not stay there. It gets exported to departments across the country. Matthews is not Atlanta. Not yet.

What The Chief Said

Police Chief Evans mentioned in his June 22nd presentation that he had visited the police chief at UNC Charlotte specifically to learn about AI surveillance technology. He offered this as a credential.

What he did not say is that UNC Charlotte and CPCC are simultaneously developing an AI behavioral detection system for CATS transit cameras. Already running on 40 cameras across the CPCC campus. Funded by the National Science Foundation with $400,000 for expansion and $550,000 for commercialization. Designed to watch riders, identify what the algorithm has been trained to call suspicious behavior, and alert CMPD.

Not facial recognition, the defenders will tell you, as though that distinction resolves the matter. As though an algorithm trained on decades of data from one of the most over-policed transit systems in the Southeast will distribute its suspicion evenly across every race and every class and every person who had the misfortune of riding the bus through the wrong neighborhood at the wrong time.

The CPCC campus is the same campus where a $116 million police training facility is being built approved without a public hearing, funded with education dollars, without a single motion from the county commissioners who approved the spending to ask what exactly was being constructed. That facility trains officers in drone operations and mock urban environments. The AI system being developed at the same institution will surveil the transit riders those officers police. The cameras Chief Evans wants downtown will watch the streets those officers patrol.

Tell me these are coincidences. I am waiting.

On The Command Center In Town Hall

Chief Evans mentioned (as a selling point) that Matthews Town Hall already has a state-of-the-art training facility upstairs that can be converted into a surveillance command center. He said they tested it during the Charlie Kirk event at Stumptown Park. He said the visual feedback was very impressive.

A surveillance command center. Already operational. Already tested on a political gathering in a public park. Inside the town hall of the municipality hosting Cop City. Nobody asked a follow-up question.

This is worth sitting with. The infrastructure for surveilling public political activity already exists in Matthews. It has already been used. The proposal before the board was not to build something new. It was to expand something already running.

What Stopped It Temporarily

The camera purchase is on hold because the vendor (almost certainly Verkada, an AI company based in San Mateo, California) wanted Matthews to litigate any disputes in California and wrote indemnity language the town attorney could not accept. The mayor said if they won't hold the price maybe we need a different vendor. The item vanished from the July 13th agenda without explanation.

Do not mistake this for a victory. The command center in town hall did not vanish. The training facility at CPCC did not vanish. The AI behavioral detection system running on forty campus cameras did not vanish. CMPD's Connect Charlotte program (which already has 1,363 business cameras integrated for live-stream access at its Real Time Crime Center) did not vanish. The hyperscale data centers approved in Charlotte's historically Black crescent neighborhoods without public hearings, the physical infrastructure on which this entire surveillance ecosystem depends, did not vanish.

One vendor's contract terms failed. The system they were meant to plug into did not. They will find another vendor. They will find better contract terms. They will add it to an agenda without telling anyone, hold the vote before opposition can organize, and thank the one person who shows up to object before moving to the next item.

The woman who drove to that special meeting on the eve of a holiday was right about what she saw. The board thanked her and moved on to the California choice of law provision. The question is whether enough of us say what she said (loudly, in public, before the next meeting) before the cameras go back on the agenda. They will go back on the agenda.