Oakland Privacy

Oakland Privacy is a citizen coalition in Oakland, California that organises against — and writes the local law that constrains — government surveillance equipment, including the algorithmic systems built on top of that equipment. It is one of the longest-running grassroots groups in the U.S. doing community control of police surveillance work, and the group whose campaigning produced the Oakland Privacy Advisory Commission, the city's standing civilian oversight body that Governing magazine called the first standing civilian privacy oversight body of its kind in the U.S. Its work spans surveillance equipment generally (cameras, automated licence plate readers, drones, ShotSpotter), the algorithmic systems trained on or layered over that equipment (face recognition, predictive policing analytics, voice and gait recognition), and the legislative and oversight scaffolding that decides whether any of those technologies can be acquired and used by city agencies in the first place.

Origin

The group emerged in July 2013 as the Occupy Oakland Privacy Working Group, formed by activists who had stayed with the anti-militarisation and anti-surveillance threads of Occupy Oakland after the 2011-2012 occupation phase wound down. Its precipitating cause was Oakland's Domain Awareness Center (DAC) — a citywide surveillance fusion centre that the city had begun building quietly under a Department of Homeland Security port-security grant, with proposed Phase II expansion that would have integrated more than 700 cameras across schools and public housing, facial recognition software, automated licence plate readers, and roughly 300 terabytes of storage for the resulting data. Working in coalition with Lighthouse Mosque, ONYX Organizing Committee / Anti-Police Terror Project, and Justice for Alan Blueford, the working group ran a campaign that culminated in the 4 March 2014 city council vote restricting the DAC to the Port of Oakland only — stopping the citywide expansion. The group renamed itself Oakland Privacy in 2016 and has continued as a standing working group with a public meeting cadence rather than re-forming around each new fight.

Strategy and tactics

Oakland Privacy's distinguishing move is to convert the public outrage of a specific surveillance fight into standing oversight infrastructure that constrains the next fight before it begins. The pattern recurs:

• The Surveillance Equipment Ordinance (later renamed the Surveillance and Community Safety Ordinance), passed by Oakland City Council in 2018, requires city agencies to seek council approval — preceded by a public Privacy Advisory Commission review — before acquiring or using any surveillance equipment. The ACLU of Northern California's account frames this as the structural inversion that the DAC fight made possible: "having the public conversation at the beginning, about whether this is appropriate or not for the community" instead of after equipment is already in place.
• The Privacy Advisory Commission itself, created by city ordinance in 2016, is the standing civilian body that conducts those reviews. Oakland Privacy members have served on the commission since its first meeting in July 2016, including longtime chair Brian Hofer, who served on the commission from its inception until his resignation in November 2025.
• Once the ordinance and commission are in place, additions to the underlying prohibition list happen by amendment rather than by case-by-case fights. The July 2019 facial recognition amendment made Oakland the third U.S. city to ban government use of face surveillance, after San Francisco and Somerville, Massachusetts. The 15 December 2020 amendment package added first-in-the-nation prohibitions on predictive policing analytics, voice recognition, gait recognition, and distance-based biometric identification — making Oakland at that point the largest U.S. city to ban predictive policing and the first to ban government use of voice recognition.

This strategic shape — coalition campaign → standing ordinance → standing oversight body → incremental amendment — is what Governing referred to when calling the Privacy Advisory Commission a national model, and what underlies the group's broader Bay Area replication work in Berkeley, BART, Davis, Palo Alto, and San Pablo. The model has shared lineage with the American Civil Liberties Union's Community Control Over Police Surveillance (CCOPS) template that has been adopted by more than twenty U.S. jurisdictions since 2016, but Oakland Privacy's contribution sits in the local infrastructure that gives those templates teeth: a citizen body that actually reviews acquisitions, with members who carry institutional memory across multiple campaign cycles.

Algorithmic accountability work

The algorithmic-systems portion of Oakland Privacy's work is what places this entry inside the make-AI-good corpus rather than alongside more general civil-liberties groups. Three specific bans, all delivered through the standing ordinance route, are load-bearing:

• Face recognition (2019). The Oakland prohibition, adopted by unanimous council vote in July 2019, bars city agencies from acquiring, obtaining, retaining, requesting, or accessing face recognition technology. The Privacy Advisory Commission's recommendation cited demographic disparities in the technology's error rates as among its grounds.
• Predictive policing (2020). The December 2020 amendments banned the city's use of predictive policing analytics. Secure Justice's account of the legislative record articulates the Privacy Advisory Commission's core rationale: that algorithms trained on historically skewed policing data direct officers back to communities they have already over-policed, compounding the bias in a self-reinforcing loop the city could not adequately mitigate.
• Voice recognition, gait recognition, and other distance-based biometric identification (2020). The same December 2020 amendment package added these prohibitions, making Oakland the first U.S. city to ban government use of voice recognition and the first to prohibit identification based on "physiological, biological, or behavioral characteristics ascertained from a distance." Brian Hofer's separate nonprofit Secure Justice records the legislative shape of the package and its first-in-the-nation framing.

The group's work on ShotSpotter, drones, and automated licence plate readers extends the same accountability frame to surveillance equipment that, while not always algorithmic in the public-discourse sense, is layered with AI-grade classification systems and feeds into the same predictive-policing-style data pipelines the December 2020 amendments were designed to cut off.

Place in the movement

Oakland Privacy is the corpus's first entity capturing the city-level community-control-of-police-surveillance organising form, complementing the regional and national pattern visible in MediaJustice's coalition work. Within the broader make-AI-good shape the corpus is mapping, the group represents a path that is materially different from both the AI-safety / pause register and the lab-pressure register: a long-running, locally embedded coalition whose theory of change runs through municipal legislation, civilian oversight bodies, and the slow accumulation of named prohibitions on specific technologies. Its bans on face recognition, predictive policing, and biometric identification at distance are among the earliest and clearest U.S. municipal-government decisions to treat specific AI systems as out-of-bounds rather than as default-permissible-with-procurement-review, and the standing ordinance-plus-commission structure it produced is the institutional shape that other U.S. cities considering similar prohibitions have most often copied.