San Diego State University quietly wired its campus with more than 1,300 AI-enabled cameras, and the students living there found out from their own newspaper, not the administration. According to Hacker News, University Police finished the $1.3 million rollout in 2024, and the full picture only surfaced after investigative journalism students at The Daily Aztec forced the camera locations into the open with a public records request.
This is significant because the cameras aren’t ordinary CCTV. They’re Avigilon units, and the manufacturer’s own site advertises facial recognition, license plate reading, behavior analysis, crowd density tracking, and audio detection. SDSU bought hardware that can identify who you are, read your plates, and analyze how you move.
What the records show
The numbers tell you who the system was built to watch:
- 1,300+ AI-capable cameras across classrooms, bookstores, dining halls, gyms, parking, and dorms
- 330+ cameras pointed at student housing, roughly 28% of the whole network
- 79 cameras on Huaxyacac alone, the largest first-year dorm
- 36 on Tenochca, 33 on Chapultepec
- 18 of 24 residential buildings on the location list
The license agreements students sign before move-in mention none of it.
The policy problem
What stands out here is the conflict with CSU’s own rules. System policy says cameras belong in public areas with no reasonable expectation of privacy, and explicitly forbids aiming them at residence halls. More than 330 now sit in exactly those buildings.
Campus officials say students are notified through the Guide to Community Living handbook and the housing website. Neither document says a word about the AI behind the lens. Campus police framed the network as little more than motion detection. “To be clear, they are not used for behavioral tracking, profiling or facial recognition,” wrote public information officer Amanda Stills, who added the school limits features on purpose and has no plans to buy more. The Avigilon contract language, per Hacker News, describes intent that reaches well past basic maintenance.
The school also declined to post signs. Stills said marking camera locations would “jeopardize public safety.” So students get constant recording with no notice at the point of recording, left to assume they’re always on camera and never sure when the AI is reading them.
Why it matters for the AI industry
The gap here is the whole story: capability versus stated use. A camera that can run facial recognition is still a camera that can run facial recognition, no matter what a policy says today about leaving the feature off. Policies change. Administrations change. The hardware stays on the wall.
This is also a pattern, not a one-off. Among all CSU campuses, only SDSU and CSU Northridge have switched on AI-powered cameras so far, but the trend runs nationwide:
- Michigan State hired a contractor to build a system that can detect barrier breaches, track individuals across campus, count crowds, and read plates
- Vendors like ZeroEyes, Flock Safety, and Volt AI are landing on campuses across the country
For anyone working in AI, this is the deployment side of the technology that rarely makes the launch announcements. The models get cheaper and better, the cameras get smarter, and the disclosure does not keep pace. The capability ships first and the conversation about consent comes later, if at all.
What to watch next
Expect more records requests. The Daily Aztec showed how effective they are, and student journalists at other schools now have a playbook. Watch for whether SDSU adds signage or amends its housing agreements, and whether CSU revisits its own privacy policy now that the contradiction is public.
The practical takeaway for practitioners and institutions: if you deploy AI-capable surveillance, the dormant features are part of what you’re accountable for, not just the ones you’ve turned on. Promises about keeping capabilities switched off have a short shelf life. Full details are available at the original source.