news-tldr.com

The important facts, the open questions, and the sources — without the churn.

← Latest briefing

Technology

Flock Safety's police AI search tools allow person tracking with limited guardrails, code analysis shows

An analysis of the surveillance vendor's software reveals natural-language search and automated watchlist tools that flag some sensitive queries without preventing them.

Updated Sep 3, 2026 · 10:18 UTC1 reports

The short version

  • A code analysis of Flock Safety software found tools enabling police officers to search and track individuals across video feeds using natural-language descriptions or automated geographic watchlists.
  • Flock's server-side moderation system blocks certain categories like race but permits searches touching on political, social, and cultural expression after an officer acknowledges a warning.
  • Civil liberties experts caution that automated moderation of search terms at scale is inherently unreliable, while Flock maintains its technology incorporates clear usage safeguards.
  • Flock plans to make stricter controls, including mandatory case codes and automated auditing, mandatory across all client departments by the end of the year.

Key facts

  • Flock Safety's surveillance software includes tools that allow police officers to search for individuals across video feeds using written descriptions and set automated watchlists over specific geographic zones.[Wired]
  • An examination of Flock's code revealed that while the system evaluates search text against sensitive categories and blocks some terms like race, queries categorized as political, social, or cultural expression only trigger a warning that officers can bypass by acknowledging it.[Wired]
  • According to reporting cited from The Washington Post, at least 50 U.S. police officers have recently faced charges or accusations over misusing license plate readers, with 46 of those cases involving Flock systems.[Wired]
  • Flock announced policy and technical changes in August that shorten default data retention periods, require case codes for searches, and introduce automated auditing, aiming to make these controls mandatory by year's end.[Wired]
  • Flock stated in response to inquiries that its tools are built to help police find relevant information while maintaining safeguards, though it did not detail the internal training instructions or mechanics of its AI models.[Wired]

What remains uncertain

  • The exact internal architecture, training data, instructions, and error rates of Flock's server-side AI screening models remain undisclosed by the company.[Wired]
  • It remains unverified whether Flock's servers will actively prevent searches submitted without valid case numbers once the company's planned end-of-year mandates take effect.[Wired]

Sources