The contract called for 400 license-plate-reader cameras across the city, paid for through the state's organized-retail-theft grant, according to the city (SF.gov). By the June 2024 announcement, SFPD and its contractor had installed a quarter of that total, with the remaining 300 expected to be operating within weeks.

Whether the network ever reached its full 400-camera footprint, and how many cameras run today, isn't addressed in the materials examined for this article.

Officials also described the license-plate readers as one piece of a larger buildout. The city said it planned to eventually connect the ALPR network with drone and camera programs authorized under Proposition E, the ballot measure voters approved in March 2024, backed by a proposed $3.7 million line in that year's budget (SF.gov).

Every example of camera-assisted policing in the materials reviewed for this article traces back to a single mayoral press release. A woman wanted on a no-bail warrant for organized retail theft was picked up on a camera in the Mission District and arrested. A vehicle tied to a carjacking near SF State was spotted, leading to three arrests. A car used in a robbery crossed into San Francisco from Oakland, was flagged by multiple cameras, and its driver was taken into custody in the Bayview (SF.gov).

The same release credited the cameras with helping sustain a citywide crime drop through May 2024, compared with the same stretch in 2023: property crime down 33%, violent crime down 13%, homicides down 38%, according to the city. Police Chief Bill Scott called the cameras "a massive help" to the department (SF.gov).

None of that is independent verification. The arrest examples and the crime statistics both come from the mayor's office, and the release doesn't separate Flock's contribution from other enforcement activity underway at the same time. A camera generating a lead that officers act on isn't proof the camera caused the arrest.

The city has said its ALPR network doesn't use facial recognition, doesn't generate alerts containing personally identifiable information, and isn't used to issue red-light or speed citations, according to a March 2024 release (SF.gov).

Those are limits on what triggers an alert, not a description of what the system stores. The public materials reviewed for this article don't specify how long a plate scan and its timestamp persist, who beyond SFPD can query one, or whether Flock aggregates data across the agencies that license its system. That's a gap in what's public, not evidence of misuse.

San Francisco's network sits inside a much larger one. The city said in March 2024 that more than 5,000 law enforcement agencies nationwide use Flock Safety systems, including dozens of agencies in the Bay Area, among them San Mateo, Alameda County, Concord, Richmond, Palo Alto, and San Jose (SF.gov).

That regional reach isn't hypothetical. In June 2024, San Jose police asked neighboring agencies for help locating a sexual-assault suspect; the vehicle turned up on San Francisco cameras near Golden Gate Park, and the suspect was arrested, the city said (SF.gov). It's the clearest documented case of San Francisco's cameras working across jurisdictional lines, though the record doesn't explain what rules govern how broadly agencies can query each other's data.

San Francisco's ordinance, Chapter 19B, has required city departments to disclose their use of surveillance technology every year since 2019 (SF.gov). The Controller's audit found that disclosure process treats all surveillance technology the same, with no mechanism to match the level of scrutiny to the risk a given system poses. It doesn't name Flock, but the finding raises an obvious question about how a regionally connected, 400-camera network gets reviewed under a law that can't distinguish it from something far smaller.

"The City passed a law to protect residents' right to privacy and safeguard their personal data. Where the law applies, it treats all technologies the same, which is a missed opportunity for stronger oversight of high-risk technologies," said Controller Greg Wagner (SF.gov). "I think we're making a strong case that the law can and should be revised in the interest of the public. As a city, we have a duty to ensure responsible use of the data we collect."

The audit made six recommendations, and the Controller's Office committed to revisiting their status every six months (SF.gov). Which of those six apply directly to license-plate readers, and which office, SFPD, the Board of Supervisors, or the Controller itself, owns implementing them, isn't specified in the materials reviewed here. SFPD, Flock Safety, the Board of Supervisors, and privacy advocates who track the ordinance don't have documented positions in the sources examined for this article.

San Francisco's ordinance has already changed once since the audit was drafted. A January 2025 update to Chapter 19B added a requirement for the city to publish a biannual Surveillance Technology Inventory report every March 1 and September 1, tracking the status of every technology on the books (SF.gov). The latest report was published September 1. Its specific entry on Flock, including current camera counts or any notes tied to the Controller's findings, isn't part of the material examined for this article.

The Controller's Office's next scheduled check on the audit's six recommendations falls roughly a year after the report's October 2025 release, putting it around this fall. Departments will file another Surveillance Technology Inventory report by March 1, 2027.

Until one of those checkpoints produces something specific about Flock, the ordinance the Controller flagged as unable to distinguish a high-risk system from a routine one remains the law governing 400 cameras and everything else on San Francisco's surveillance inventory.