The County of Santa Clara Government Center tower in downtown San Jose lit up at night

Inside Santa Clara County’s Surveillance Ledger: 93 Technologies, 66 Million Plate Scans and the Audit Gaps Supervisors Will Review Tuesday

, ,

·

Editor-in-Chief

·

PREMIUM

Top image: The County of Santa Clara Government Center in San José, where the Board of Supervisors meets. Photo: Xnatedawgx / CC BY-SA 4.0 via Wikimedia Commons

When the Santa Clara County Board of Supervisors meets Tuesday, Oct. 20, at 9:30 a.m., one of the longest documents in its packet will draw little debate. Item 46 asks the board to “receive” the county’s Annual Surveillance Report for the fiscal year that ended June 30, 2026, an informational filing required under the county’s decade-old surveillance ordinance. No vote on policy is attached.

But read in full, the report and its two appendices, which together run to dozens of pages, are the most detailed public accounting of how a large California county watches, records and shares information about the people in it. They also put on paper, for the first time, what the county’s controversial contract-city license plate readers actually did before supervisors cut them off in February.

This analysis is based on the county’s report by Deputy County Executive Megan Doyle, Appendix A on higher-risk technologies, Appendix B on medium- and low-risk technologies, and a separate District Attorney request (Item 59) on the same agenda.

The inventory

The county classified 93 surveillance technologies as of June 30: 48 low-risk, 23 medium-risk and 22 higher-risk. Under a February 2025 amendment to the ordinance, the County Executive, after consulting County Counsel, sorts each technology by factors including privacy impact, civil liberties concerns, security vulnerabilities and data-sharing risk. Higher-risk technologies need a board-approved surveillance use policy; lower tiers move through a streamlined process.

The Office of the Sheriff accounts for 15 of the 22 higher-risk technologies. The District Attorney has three, Probation two, the Public Defender and Alternate Defender one, and one is shared among Pretrial Services, Probation and the Sheriff.

Three technologies were downgraded during the year. A DA access-card system moved from higher to medium after the county confirmed it stores a “biometric-derived mathematical hash” rather than fingerprint images; the earlier rating, the report says, was based on “the incorrect assumption that full fingerprint images were retained.” Jail telephone monitoring was reclassified based on verbal disclosures and signage telling callers that calls may be recorded. A Sheriff screening system was downgraded after the county confirmed it captures images “only of screened items, not of individuals.”

The plate readers, in numbers

The most politically charged entry is No. 20 in Appendix A: automated license plate readers the Sheriff’s Office operated for the City of Saratoga, the City of Cupertino and the Town of Los Altos Hills, which contract with the Sheriff for policing. The cameras captured plate numbers and “vehicle identifiers such as color, make, model, and distinguishing features (e.g., bumper stickers, tow racks),” the appendix says.

For July 2025 through February 2026, the report lists:

  • Cameras: 22 in Cupertino, 32 in Los Altos Hills and 56 in Saratoga.
  • Plate scans: 38,699,901 in Cupertino, 6,088,767 in Los Altos Hills and 21,162,945 in Saratoga, about 66 million in all.
  • Results in Cupertino: seven stolen vehicles and three stolen plates recovered, nine arrests and five warrants served.
  • Results in Saratoga: two stolen vehicles recovered, three arrests and one warrant served.
  • Results in Los Altos Hills: “investigative leads only.”

The report states that no ALPR data was shared with outside entities and that the Sheriff ran biweekly and monthly audits checking user access, case numbers and search criteria. It concludes the technology “was effective in identifying and locating stolen vehicles.”

The counts end in February for a reason. On Feb. 24, as the appendix notes, the board directed the Sheriff’s Office to stop supporting the contract cities’ ALPRs. KQED reported at the time that supervisors voted 3-2 to bar the Sheriff from using data from Flock Safety cameras, which all three jurisdictions had relied on, amid concerns about federal immigration agencies’ access to Flock data elsewhere. “Flock is a problematic company,” Supervisor Betty Duong said then. Supervisor Susan Ellenberg voted no because she viewed plate readers generally as an “excessive invasion of privacy,” KQED reported. The annual report itself does not name the vendor.

The figures allow a rough ratio that neither side in that debate had in the public record: across the three jurisdictions, the report lists 12 arrests, six warrants served, nine stolen vehicles and three stolen plates recovered from about 66 million scans over roughly eight months, plus unquantified “additional investigative leads.” That is not a measure of deterrence or of cases where a plate hit was one piece of a larger investigation, and the report does not attempt one. The county’s own separate ALPR policy, entry No. 21, records “no operational use” during the year.

The issue is not limited to the South Bay’s smaller cities. The City of San Mateo recently switched vendors and narrowed sharing rather than abandoning plate readers, and a judge recently upheld San José’s camera network against a legal challenge.

Cameras on deputies, and features switched off

The Sheriff’s body-worn and dash cameras generated 725,195 recordings totaling 103,535 gigabytes, about 58,133 hours, during the year, according to Appendix A. The office maintains 1,362 body cameras for about 1,276 personnel and keeps “partner agency” sharing capability with 20 outside law enforcement agencies; 201 dash-camera cases or files were shared externally, many with the county’s independent oversight office.

One detail looks ahead. The Sheriff upgraded to Fleet 3 in-car cameras, which support live streaming and automatic license plate reading. “Those functions are disabled,” the appendix says, and the Sheriff plans to evaluate them in the future, “at which time an updated surveillance use policy will be submitted to the Board of Supervisors for approval.” In other words, the hardware for a new layer of plate reading already sits in county patrol cars; switching it on would require a public vote.

Phones, cars and forensics

The DA’s office and its REACT high-tech crime task force completed at least 151 forensic examinations of phones, computers and other devices, plus 258 at the Crime Laboratory, during the year, and shared data with agencies including the FBI, IRS Criminal Investigation, the U.S. Postal Inspection Service, the Napa County DA and the Thurston County Sheriff’s Department in Washington state. The appendix notes that “the forensic tools do not include built-in audit features,” so supervisors rely on credentials, activity logs, monthly activity sheets and periodic audits. They found no violations.

On the same agenda, Item 59, District Attorney Jeffrey F. Rosen asks the board to approve new tools from Berla, which the memo describes as technology to “extract and analyze digital evidence from modern vehicle infotainment and telematics systems.” The ordinance requires the board to find that benefits outweigh costs, “including both the financial costs and reasonable concerns about the impact on and safeguards for privacy, civil liberties, and civil rights.” The memo’s stated reasoning: “the benefits of the surveillance technology outweigh the costs because there is no cost to the Office of the District Attorney,” as the purchase will be reimbursed by the state through restricted REACT funds. The memo adds that the office is “sensitive to civil liberty concerns” and conducts frequent audits.

The Public Defender’s office, by contrast, reports using its Cellebrite extraction equipment 27 times, and says the extracted data often provided exculpatory or mitigating evidence for clients.

Where the rules were broken

The report lists three known violations, up from two the year before but down from 13 in fiscal 2023-24. Two were the same problem in two Sheriff systems: X-ray images of bags and packages placed on conveyor belts “were being retained for approximately 180 days,” though the policy “only authorizes real-time viewing of scans and does not permit the retention of any scanned images.” The fix was not to delete the images but to set a 90-day default retention period and to update the policy “to formally incorporate this authorized retention period.”

The third involved the Social Services Agency’s camera systems at transitional shelter care facilities for children: records went to the California Department of Social Services under an existing agreement, but CDSS was not listed as an authorized recipient in the policy. “No data breach occurred,” the report says.

Two of the three violations were found during the Privacy Office’s own assessments, which rose to 16 from 11 the prior year. The report records no community complaints about any surveillance technology during the year.

What else is watching

Appendix B fills in a long tail: a new Sheriff language-translation platform rated medium-risk; a Parks and Recreation subscription to CityData, which visualizes visitation patterns using anonymized location data from mobile devices, rated low-risk; and Probation’s new full-body scanners at Juvenile Hall and James Ranch, which ran 3,533 scans of 280 youths in their first six months. Probation’s own analysis “did not show a significant reduction in contraband incidents,” although the department says the scanners standardized intake and will keep collecting data.

The long view

Santa Clara County’s ordinance was an early attempt to make surveillance a matter of public record rather than procurement. The 2026 report shows both what that system does well, a full inventory, published numbers and self-reported violations, and where it is thin. Effectiveness sections are written by the departments that use the tools and nearly always conclude they work. Some of the most sensitive tools lack built-in audit trails. Cost-benefit findings can rest on the observation that the state is paying.

The plate-reader numbers are the clearest example of why the reporting matters. For years the debate over Flock in Cupertino and Saratoga was carried on in principle. Now there is a baseline: tens of millions of scans, a few dozen enforcement outcomes and a stack of leads. Whether that is a reasonable return is a value judgment, not a technical one, and the next test will come when the Sheriff asks to turn on the plate-reading features already installed in its patrol cars.

Sources

Leave a Reply