Facial recognition systems are scanning shoppers as they enter a growing number of Grocery Outlet stores in the San Francisco Bay Area, comparing their faces against watchlists of people suspected of theft, violence, or other unlawful conduct.
Signs disclosing the use of biometric face-matching software have been documented at Grocery Outlet stores in Pleasant Hill and Concord, as well as four San Francisco locations in the Mission, Portola, Bayview, and Richmond districts.
The signs direct customers to the privacy policy for SAFR Guard, the system being used by the stores.
The full extent of the deployment remains unclear. Grocery Outlet operates through a network of independently operated stores, and neither the company nor local store managers have publicly explained whether the installations are part of a corporate program or decisions made by individual operators.
Grocery Outlet did not respond to requests for comment from several news organizations.
SAFR Guard is an AI-powered security solution for the retail loss prevention and asset protection market to prevent shoplifting and retail crime before it happens.
SAFR Guard is a product of SAFR LLC, a subsidiary of RealNetworks.
Cameras at participating stores capture facial images as customers enter and create encrypted biometric templates inside the camera. Those templates are compared with an in-camera watchlist. A potential match is sent for additional verification before the retailer receives an alert.
Retailers provide SAFR with images and descriptions of incidents involving people they suspect of illegal conduct. SAFR says it reviews and approves submissions before they are added to a watchlist.
The system may also share selected records concerning suspected violence or high-dollar thefts with other participating retailers in the same geographic area.
SAFR says the facial recognition template generated for a shopper who does not match a watchlist is held inside the camera for no more than one second and then deleted.
Its privacy policy, however, says encrypted images of people who are not designated as subjects of interest can be retained for as long as 30 days. Camera footage is also stored locally for 30 days so retailers can review incidents and create watchlist records.
Images and incident information concerning people designated as subjects of interest can be retained for up to 12 months. Some facial recognition records may be kept longer if SAFR reviews them and determines continued retention is necessary.
SAFR President Charisse Jacques said the system “is not a law enforcement system and has no connection with law enforcement” and does not maintain a general purpose database of the public.
She said the company has never shared retailer or shopper information with a government agency, including Immigration and Customs Enforcement, and has not received a subpoena or court order demanding the information.
The company’s privacy policy nevertheless permits disclosure when required by a court order, law, or other legal process.
The technology has divided Grocery Outlet customers.
The deployment comes as California retailers continue to confront elevated shoplifting levels, although the most recent statewide figures do not support claims that theft is still increasing.
Shoplifting declined by about 9 percent in 2025, according to the Public Policy Institute of California, but remained approximately 36 percent above the 2019 level.
California’s overall property crime rate fell to its lowest point in four decades.
Privacy advocates say the central problem is that the system must examine everyone’s face to find the relatively small number of people on a watchlist.
“This is a dragnet that scans everyone. Even if you’ve done nothing wrong, your face is being scanned,” said Electronic Frontier Foundation staff attorney Mario Trujillo.
Critics also question how retailers establish that an incident occurred, whether employees may mistake lawful behavior for theft, and what process is available to challenge an incorrect watchlist entry.
The Federal Trade Commission (FTC) barred Rite Aid from using facial recognition surveillance for five years after alleging that its system produced thousands of false matches and caused employees to follow, search, or publicly accuse customers who had done nothing wrong.
The FTC said false-positive matches occurred more frequently in stores located in predominantly Black and Asian communities.
San Francisco’s prohibition on government facial recognition does not apply to private retailers.
SAFR says it requires participating businesses to post signs disclosing the technology, but reports from the stores found that some notices were small, difficult to read, or overlooked by customers entering the building.
Article Topics
biometric matching | biometrics | California | facial recognition | retail biometrics | SAFR