AI Model Wiki

    Skip to main content
    12 min read
    Lesson 3 of 6

    Facial Recognition: Capabilities & Ethics

    From the course AI for Law Enforcement & Public Safety

    Facial recognition technology (FRT) is one of the most debated AI applications in law enforcement. Understanding both its capabilities and limitations is essential.

    How Facial Recognition Works

    1. Detection: AI identifies faces in images or video feeds.
    2. Feature extraction: The system maps facial geometry — distance between eyes, jawline shape, nose contours — into a mathematical representation (faceprint).
    3. Matching: The faceprint is compared against a database of known individuals.

    Law Enforcement Use Cases

    • Suspect identification: Matching surveillance footage against mugshot databases.
    • Missing persons: Scanning crowds or social media for missing children or vulnerable adults.
    • Identity verification: Confirming identities at borders or secure facilities.
    • Cold case resolution: Matching old evidence photos against current databases.

    Accuracy Concerns

    Studies by NIST and others have shown that many facial recognition systems perform worse on women and people with darker skin tones. This disparity can lead to wrongful identifications. Agencies should:

    • Only use systems that have been independently tested for demographic accuracy
    • Require human verification before any action based on a match
    • Establish clear policies on when and how FRT can be used
    • Engage community input on FRT deployment decisions

    Legal Landscape

    Several cities and states have enacted bans or moratoriums on government use of facial recognition. Officers should understand local regulations and the evolving legal framework.

    Key Takeaways

    • •Facial recognition maps facial geometry for identity matching
    • •Accuracy varies significantly across demographics — bias testing is critical
    • •Many jurisdictions restrict or ban law enforcement use of FRT
    • •Human verification should always follow an AI match
    0 of 6 lessons marked complete

    Progress is kept in this browser only.