UNC's Flock cameras under debate: 23 AI license plate readers and a $70,000 contract that expires in February

By The Daily Tar Heel · November 3, 2026.

The University of North Carolina at Chapel Hill (UNC) has 23 Flock cameras installed (automated license plate readers that run AI software), and its contract expires in February 2027. According to the university, the current contract began on February 21, 2024, and costs $70,000 a year. Students, advocacy groups and faculty are now asking whether renewing it is worth it.

The cameras, made by Flock Safety, photograph passing vehicles. The software identifies the license plate and the car's features. According to the company's own website, authorized users such as police can search those records for vehicles linked to an investigation. UNC signed an initial agreement in September 2023 and canceled it a month later over licensing issues.

A university spokesperson argues that the technology helps identify vehicles linked to criminal activity. The spokesperson points to its value in urgent cases, such as missing or endangered people and Amber Alerts (alerts issued when a child is abducted). According to the spokesperson, UNC Police do not share information with other law enforcement or federal agencies and keep the data for 30 days. On October 5, Police Chief Brian James told the Faculty Executive Committee that automatic data sharing with other North Carolina agencies had been ended. Those agencies must now request the data from UNC.

Opposition is coming from several directions. First-year student Gray Crook wants evidence that the cameras improve safety. Nadia Mokhtarzada Blake of TransparUNCy warns that "just because the information isn't shared initially doesn't mean the policy can't change." Her organization is calling for the contract not to be renewed. Hannah Hayes, co-chair of Sunrise UNC, argues that real safety means investing in students, not in surveillance. Hayes cites the recent racially motivated attack at Greenlaw Hall and says her group is exploring alternatives. She also says the police chief has been open to dialogue.

The legal questions remain unresolved. Jeff Welty, a professor of public law at the UNC School of Government, says the key question is whether these cameras, alone or combined with other tools, build such a complete picture of someone's movements that it amounts to a search under the Fourth Amendment (which protects against unreasonable searches). So far, courts have sided with Flock, but Welty warns that this guarantees nothing as the network grows. He separates that constitutional question from possible misuse by officers, which he believes is a matter for training, policies, audits and penalties. His colleague Kristina Wilson adds that it is unclear whether the images count as public records.

Our take: the debate at UNC isn't really about a camera. It's about who sets the rules for AI surveillance infrastructure that is becoming cheap and routine. The university's case is reasonable for urgent situations, but the burden of proof lies with whoever does the watching. Before renewing, UNC should publish the specific cases the 23 cameras helped solve. If it has them, that is its strongest argument. If not, Crook's question remains unanswered.

Blake's warning deserves more weight than it usually gets. The current limits on use are internal policy decisions and can be reversed without public debate. In general, AI-based surveillance tends to expand once it is in place. That is why limits on retention, sharing and auditing should be written into binding, verifiable rules or contracts, not just public statements.

There are reasons not to be alarmist. License plate readers are a limited technology, with 30-day retention according to the university, and an open debate on campus is exactly the healthy way to decide. The short-term risk is that mass tracking becomes normal without evidence that it works. In the long run, AI can deliver real safety, but only if it is built with clear limits and human oversight. A campus where students and police discuss what surveillance should "look like" is a better starting point than one where it is installed quietly.