When the Machine Accuses the Innocent

Facial recognition technology is being deployed by law enforcement nationwide. Marketed as an objective tool, it carries a hidden, systemic bias that disproportionately harms Black communities, leading to a modern civil rights crisis fueled by flawed algorithms and institutional failure.

Uncover the Truth

The Weight of Evidence

The claim of racial bias is not anecdotal. It's a scientific fact proven by landmark academic and government studies. These visualizations reveal the stark, disparate error rates at the heart of the crisis.

The "Gender Shades" Bombshell

The 2018 "Gender Shades" study exposed catastrophic intersectional bias. While error rates for light-skinned men were minimal, they soared for darker-skinned women, reaching nearly 35% for a simple gender classification task. Click the buttons below to see the failure rates for each company's algorithm.

The Government Confirms It

The U.S. government's own testing by the National Institute of Standards and Technology (NIST) confirmed the industry-wide problem. Their 2019 report analyzed 189 algorithms, finding a clear pattern of "demographic differentials."

For one-to-many searches used by police, algorithms were:

10 to 100x

more likely to return a false positive (an incorrect match) for Black and East Asian faces compared to white faces. This type of error leads directly to wrongful arrests.

The Anatomy of Bias: The "Coded Gaze"

This bias isn't an accident. It's built into the technology at every stage. Researcher Joy Buolamwini calls this the "Coded Gaze"—how human prejudices become encoded into our machines.

🗃️

1. Unrepresentative Data

Algorithms learn from data. Most training datasets are overwhelmingly composed of lighter-skinned male faces. A system trained on a skewed reality cannot perform fairly for everyone.

⚙️

2. Biased Algorithms

Early systems required engineers to manually define facial features. This replicated the human "other-race effect"—developers unconsciously optimized for features common to their own race, making the system less accurate for others.

📷

3. Biased Hardware

Film chemistry and camera sensors were historically optimized for lighter skin. This results in poorer quality, underexposed images of darker faces, robbing the algorithm of crucial data and increasing error rates.

Innocence Lost: A Chronicle of Wrongful Arrests

A "false positive" isn't a statistic. It's a knock on the door, handcuffs, and a life shattered by a machine's mistake. To date, every known wrongful arrest from facial recognition has involved a Black person. Click on a case to learn their story.

Reality Check: Industry Claims vs. Real-World Harm

The tech industry promotes a narrative of near-perfect accuracy. This narrative crumbles when compared to how the technology is actually used—and fails—in the real world.

📢 The Industry Claim

  • "Our best algorithms are over 99% accurate and show no bias."
  • "Police only use it as an investigative lead, not for arrests."
  • "Errors are caused by bad image quality, not the algorithm."

💥 The Harsh Reality

  • Lab accuracy using high-quality photos doesn't reflect grainy, real-world surveillance footage where all wrongful arrests occur.
  • Police repeatedly treat the "lead" as fact, leading to "automation bias" and tainted investigations with no other evidence.
  • A tiny error rate applied to databases of millions creates thousands of false matches, disproportionately harming people of color.

The Fight for Justice

In response to this crisis, a movement for regulation is growing. Cities, states, and civil rights groups are fighting back through legislation and litigation, creating a patchwork of protections across the nation.

Bans on Government Use

Cities like San Francisco, Boston, and Portland have enacted full bans on police use of FRT.

Warrant Requirements

States like Maine, Utah, and Massachusetts require police to get a warrant before running a search.

Limits on Use

States like Illinois and Maryland limit FRT use to only serious violent felony investigations.

A Call for a Federal Moratorium

Despite local progress, there is no federal law regulating facial recognition. Civil rights groups and lawmakers are calling for a national moratorium to halt the harm and allow for a public debate on whether this technology is compatible with a free and just society.