A longer story
This didn't start with cameras.
The cameras and algorithms feel new. The impulse behind them is not. Long before facial recognition, American cities built systems to watch, mark, and control the movement of Black and Indigenous people — and scholars trace a direct line from that history to the tools on the table today.
This isn't about guilt. It's about memory. A city that understands where a technology comes from is better able to decide, clearly and freely, where it should go.
The long arc
Lantern laws (18th century)
New York and other cities required Black, mixed-race, and Indigenous enslaved people to carry a lit lantern after dark — and deputized any white person to stop those without one. Scholar Simone Browne reads the lantern as an early “supervisory device,” a forerunner of identification and tracking.
Slave patrols
Movement was governed by a pass system and enforced by slave patrols — organized as early as 1704 in South Carolina and widely cited by historians as among the earliest forms of organized policing in America. After the Civil War, their functions were absorbed into Southern police forces.
Branding & the ledger
Browne reads the branding of enslaved people as an early biometric mark, and the 1783 “Book of Negroes” — a ledger of self-emancipating people — as “a form of biometric information technology,” a new genealogy of the modern passport.
To today
Browne's point isn't that past and present are identical, but that “some of the practices that we see happening now have earlier articulations.” Cameras calibrated to lighter skin, databases that follow people who've done nothing wrong — these are, in her framing, the newest chapter of a very old book.
The thinkers
People who have studied this for decades
Simone Browne
“Racializing surveillance” — surveillance has long produced and enforced racial order; “dark sousveillance” names the ways people resist it.
“...those moments when enactments of surveillance reify boundaries, borders, and bodies along racial lines.”Dark Matters (2015) ↗
Ruha Benjamin
“The New Jim Code” — ostensibly neutral tech re-encodes old hierarchies while being sold as more objective than what it replaces.
“...technologies that reflect and reproduce existing inequities but that are promoted and perceived as more objective and progressive.”Race After Technology (2019) ↗
Sarah Brayne
Big-data policing doesn't remove human bias — it hides discretion earlier and widens the net to people with no police contact.
“...the threshold for inclusion in law enforcement databases is lower, now including individuals who have not had direct police contact.”Predict and Surveil (2021) ↗
Chris Gilliard
“Luxury surveillance” — the same tracking feels like care when you buy it, and like control when it's imposed on you.
“The difference between a smartwatch and an ankle monitor is, in many ways, a matter of context.”“The Rise of Luxury Surveillance,” The Atlantic (2022) ↗
Virginia Eubanks
The “digital poorhouse” — automated eligibility and risk-scoring systems surveil and discipline the poor, and are tested first on those least able to resist.
“Automated systems control which neighborhoods get policed, which families attain needed resources, and who is investigated for fraud.”Automating Inequality (2018) ↗
The pattern today
The same shape, in the data
Cameras cluster where the most people of color live
Amnesty International mapped 25,500+ NYC cameras and found the higher the share of non-white residents, the higher the concentration of facial-recognition-compatible CCTV — “effectively a digital stop-and-frisk.”
Amnesty International (2022) ↗Most gunshot alerts lead nowhere
A study of Chicago's ShotSpotter found ~89% of alerts turned up no gun-related crime — roughly 40,000 dead-end deployments — concentrated in the most heavily Black and Latino districts.
MacArthur Justice Center; ACLU ↗Predictive policing feeds on itself
Feeding Oakland arrest data into a predictive system would target Black people at ~twice the rate of white people — despite similar drug-use rates — because biased data sends police back to the same blocks. A “runaway feedback loop.”
Lum & Isaac, Royal Statistical Society (2016) ↗The technology misreads the people it's aimed at
Federal testing (NIST, 2019) found face-recognition false positives for Black and Asian faces 10 to 100 times higher than for white faces. “Gender Shades” found error rates up to 34.7% for darker-skinned women vs. near-zero for lighter-skinned men.
NIST NISTIR 8280; Buolamwini & Gebru ↗Detroit, the current chapter
Detroit is the current chapter of this story. Project Green Light, launched in 2016, streams hundreds of cameras to police — clustered, by independent mapping, in the predominantly Black neighborhoods of a city that is 77% Black. In 2017 the city added DataWorks Plus facial recognition that could run against that network.
Two numbers often get confused, and shouldn't. In 2020, the police chief said the software used alone would misidentify someone “96% of the time.” Separately, of 69 facial-recognition searches DPD ran in 2021, 68 — about 98% — were of Black people. One number is about accuracy; the other is about who it's pointed at.
The cost was human: Robert Williams, Michael Oliver, and Porcha Woodruff — all Black, all misidentified, Woodruff arrested while eight months pregnant. Their cases led to a 2024 settlement the ACLU calls the nation's strongest limits on police facial recognition. Detroit didn't just inherit this history — it helped write its newest page, and then began to correct it.
Read deeper
Show up when it counts
Decisions get made in rooms.
We'll tell you which room, and when.
The most powerful thing a resident can do is be present before the vote. Subscribe and we'll email you ahead of every City Council session, Police Commissioners meeting, and public hearing that shapes our city.