Flock's gunshot detection microphones will start listening for human voices (www.eff.org)

Flock Safety—known for its networked license-plate readers—announced a new “human distress” audio feature for Raven, its acoustic gunshot‑detection microphones. Raven units already capture street-level sound and use machine learning to flag gunshots; marketing materials now show alerts for “screaming” and invite cities to apply for early access to voice‑based distress detection. Flock has not explained how the model works, whether audio is processed on‑device or streamed/stored, or what thresholds and safeguards govern alerts. The change matters because it expands surveillance scope from impulsive, transient gunshot signatures to human voices and potential conversations, raising legal, privacy, and public‑safety concerns. Acoustic gunshot systems already suffer false positives (e.g., fireworks, backfires) and prompt risky police responses; adding voice detection magnifies risks of misclassification, unlawful listening under state eavesdropping laws, broader data sharing, and mission creep—especially given Flock’s prior controversies over data access and municipal pushback. For technologists and policymakers, key technical questions are model accuracy, on‑device vs. cloud inference, retention/access policies, bias and robustness in noisy urban environments, and independent audits to prevent civil‑liberties harms before wider deployment.