Facial Recognition False Arrest

Hazard Definition

Facial recognition false arrest refers to incidents where law enforcement agencies rely on automated facial recognition systems that incorrectly match individuals to criminal suspects, resulting in wrongful arrests, detentions, interrogations, and associated harms. These misidentifications occur when algorithmic systems return false positive matches that are then acted upon without adequate human verification or investigative corroboration.

Mechanism of Harm

Facial recognition false arrests arise through a combination of technical limitations and procedural failures.

Algorithmic accuracy disparities: Independent evaluations by the NIST Face Recognition Vendor Test have documented significant accuracy disparities in facial recognition algorithms across demographic groups. Many algorithms exhibit higher false positive rates for Black individuals, women, and younger people compared to white males, creating disproportionate misidentification risk for these populations.

Low-quality source images: Facial recognition searches often use surveillance footage, social media images, or other low-resolution sources that degrade matching accuracy. Blurry images, poor lighting, partial faces, and non-frontal angles increase false match probability substantially.

Insufficient human review: When investigators treat algorithmic matches as definitive identifications rather than investigative leads requiring corroboration, wrongful arrests result. Documented cases reveal instances where officers made arrests based primarily or solely on facial recognition output without independent verification.

Documented Incident Patterns

Civil rights litigation, investigative journalism, and law enforcement disclosures have documented multiple wrongful arrest cases.

Documented wrongful arrests: At least six cases of wrongful arrest based on facial recognition misidentification have been publicly documented and reported by major news organizations. All publicly known cases as of the publication date of this entry involved Black individuals, consistent with documented algorithmic accuracy disparities.

Detention and interrogation harms: Documented victims have been held in jail for periods ranging from hours to days, subjected to interrogation, required to post bond, and forced to defend against criminal charges before misidentifications were recognized. Victims report significant emotional distress, employment impacts, and reputational harm.

Civil litigation: Multiple lawsuits have been filed against law enforcement agencies alleging civil rights violations arising from facial recognition misidentification. Some cases have resulted in settlements; others remain in litigation.

Regulatory Status

No federal law specifically regulates law enforcement use of facial recognition technology. Deployment practices, accuracy thresholds, and human review requirements vary by jurisdiction without standardization. The DOJ Civil Rights Division has authority to investigate pattern-or-practice civil rights violations by law enforcement agencies, which could include systemic misuse of facial recognition technology.

Several cities and states have enacted bans or restrictions on government facial recognition use, including San Francisco, Boston, and several other municipalities. Some states have enacted laws requiring disclosure of facial recognition use in criminal proceedings.

The absence of mandatory accuracy standards, audit requirements, or incident reporting obligations means the full scope of facial recognition misidentification remains unknown. Wrongful arrests may occur without public disclosure unless victims pursue legal action or media coverage.

Known Data Gaps

• Total number of arrests made based on facial recognition matches nationwide
• False positive rate in actual law enforcement deployments versus controlled testing
• Number of wrongful arrests that have not been publicly disclosed or litigated
• Demographic breakdown of individuals subjected to facial recognition searches