Flock Crime Prevention Tracker

Methodology

This database is assembled automatically each day. Candidate articles are gathered from the GDELT Project and Google News using searches for Flock Safety cameras in crime coverage. Each article is then classified by a large language model (Anthropic’s Claude) to determine whether it describes a concrete incident in which a Flock camera assisted in solving or preventing a crime — excluding privacy debates, procurement news, and company press releases — and to extract the date, location, crime type, the camera’s role, and the outcome. Stories covering the same incident are merged into a single entry.

Entries reflect claims made in news reports, which typically rely on police statements; inclusion here is not an independent verification of the camera’s role. Coverage is limited to English-language outlets indexed by the sources above, so the database understates the true number of such incidents. Each entry links to its source article(s). The full dataset is available as a CSV download. Spotted an error or a missing incident? File a report — submissions are automatically checked against the database and cited sources, then reviewed by the maintainer; every accepted correction is public in the edit history.