Grocery Outlet stores in San Francisco and the East Bay are using facial recognition technology to identify suspected shoplifters, raising concerns about surveillance and consent.

A Grocery Outlet location on South Van Ness Avenue in San Francisco on July 16, 2026. (Beth LaBerge/KQED)

This week, Lee Hepner rode his bike to the Grocery Outlet in the Mission District in San Francisco — but not to go shopping. 

He was looking for a postcard-sized sticker on the front door that some might miss, alerting customers that the store was scanning the face of every person who walked in.

“Face Matching software is being used to prevent shoplifting,” the sign read, accompanied by a QR code with information about opting out of your data.

“Rarely do I have a situation where surveillance technology is popping up in my backyard,” said Hepner, senior legal counsel at the American Economic Liberties Project. In 2019, he helped write policy making San Francisco the first city in the country to bar its own agencies, including police, from using facial recognition. 

Inside, he counted three cameras near the entrance and dozens more down every aisle and over every checkout. 

A sign on the door notifies customers about facial matching at a Grocery Outlet location on South Van Ness Avenue in San Francisco on July 16, 2026.

“It is just blanketed with surveillance cameras,” he said.

Emeryville-based Grocery Outlet has rolled out software by an AI company called SAFR Guard at several Bay Area stores, including Pleasant Hill and the San Francisco location Hepner visited. 

The technology is designed to identify people suspected of shoplifting, according to its privacy policy, and notifies employees if someone on a watchlist enters a store. SAFR Guard builds a database of “subjects of interest” across the retailers using it, Hepner said — meaning whoever lands on a watchlist at one store can follow them to another.

It’s part of a quiet spread of automated anti-theft tools into neighborhood groceries and retail businesses. As retail theft rates have increased in recent years — shoplifting in California was 47.5% higher in 2024 than before the pandemic, according to the Public Policy Institute of California — companies like Target, Walmart and grocers like Wegmans have introduced the software. 

Hepner said San Francisco banned the technology for government use precisely because research shows it “disproportionately misidentified people with darker complexions and faces of women.”

SAFR Guard defends the system as narrow and privacy-conscious. In a statement, company President Charisse Jacques said the software “is not a law enforcement system and has no connection with law enforcement,” and it does not build “a general-purpose face matching database of the public.” 

A Grocery Outlet location on Bayshore Boulevard in San Francisco on July 16, 2026.

The system only holds data on people a retailer has added to a security list after an incident, such as suspected theft, Jacques said. According to their privacy policy, records on suspected individuals are kept for up to 12 months, while encrypted images and general CCTV footage are automatically deleted after 30 days.

Shoppers who aren’t matched to that list are never stored at all: their biometric data “never leave the camera and are immediately deleted.” She said that SAFR has never shared retailer or shopper data with any government agency, including ICE, and has never received a court order or subpoena for it.

Grocery Outlet did not respond to requests for comment.

Hepner doesn’t dispute that the store’s current privacy policy is narrow. His worry is what would happen if the software is abused. 

“You make policy to address the risk of harm,” he said, “even if that harm is on the fringes of the use of that technology, because I can guarantee you it will be mainstream tomorrow.”