Flock Safety Announces LPR Privacy Guardrails, But Audit Tool Remains a Black Box
Flock Safety announced a suite of new policies and tools on August 13, 2026 that the company says will curb abuse of its automated license plate reader (ALPR) systems and hold law enforcement customers accountable for misuse.
The most concrete change is a reduction in default data retention. Flock now recommends that police departments retain ALPR data for seven days rather than the previous 30-day recommendation. For cases requiring longer retention, the company introduced a feature called Evidence Mode, which can be activated when a customer needs to hold data beyond the default window and requires a specific case number to do so.
Flock also gave customers the ability to set limits on what data they share with other departments in the Flock network, gated by the type of offense under investigation. A department investigating a property crime, for instance, could restrict data sharing so that only relevant queries flow to partner agencies.
The centerpiece of the announcement, however, is a tool called Audit Assistance, which Flock launched in April 2026. The company describes it as a feature that can "surface atypical activity early and review it quickly with a clear, documented workflow." Ashley Haber, Flock's head of trust and compliance, said the tool can flag "what may look like an odd search history from a specific user." Co-founder Paige Todd offered a more specific example: it "flags unusual patterns early," such as a user searching the same license plate repetitively for more than 30 days. The tool can automatically lock out flagged users until an administrator intervenes.
Major Patrick Krieg of the Dunwoody, Georgia Police Department, a Flock customer, described Audit Assistance as "an algorithm that has notified us of any type of bias."
Flock says more than one-third of its customers have already enabled Audit Assistance and is now requiring all customers to turn it on by the end of 2026.
What Flock has not done is explain how the tool actually works. TechCrunch asked the company how the Audit Assistance algorithm functions, what data it is trained on, whether it was developed using a large language model, what patterns it is designed to surface, and what exactly gets flagged as abnormal activity. According to TechCrunch, Flock confirmed that Audit Assistance is not based on machine learning, but did not provide further technical detail on the detection logic.
That gap matters. An anomaly-detection system operating on law enforcement query logs is only as trustworthy as the rules or heuristics it encodes. Without knowing what thresholds trigger a flag, whether those thresholds are static or tunable per deployment, or what categories of query behavior fall within scope, it is difficult for an independent auditor or civil liberties observer to evaluate whether the tool would catch the forms of abuse most commonly associated with ALPR systems: officers running plates for personal reasons, tracking individuals without investigative pretext, or querying data on political or journalistic targets.
The ACLU responded to Flock's August 13 announcement the same day, arguing that the company's updates fail to acknowledge how easily users have circumvented the "search reason" checks already present in Flock's systems. If the existing audit controls are bypassable by typing a plausible justification into a free-text field, the effectiveness of any downstream anomaly detection depends on whether the new tool examines the substance of those justifications or merely the pattern of query behavior around them.
The data retention reduction from 30 days to seven is a tangible change with clear privacy implications. Shorter retention windows reduce the volume of historical location data available for retrospective queries, which limits both legitimate investigative reach and the surface area for misuse. The Evidence Mode requirement for a case number introduces a lightweight accountability checkpoint for extended retention, though its enforcement depends on departmental discipline rather than any external verification.
Flock's networked data-sharing model has been a particular flashpoint for privacy advocates. The ability for a plate read collected by one department to surface in another department's queries, across jurisdictions, expands the surveillance footprint of a single camera deployment well beyond its local context. The new per-offense sharing limits address this concern at the policy layer, giving customers granular control over what flows outward. Whether departments will configure those limits conservatively or permissively remains an open question.
Flock is operating in a space where the technology has outpaced the regulatory framework. ALPR systems are deployed widely across U.S. law enforcement agencies with no federal standard governing retention periods, data sharing, or audit requirements. State-level statutes vary considerably. In that vacuum, vendor-imposed guardrails carry weight that they would not in a well-regulated environment, and the opacity of a core accountability tool is correspondingly harder to dismiss.
The company's decision to mandate Audit Assistance across all customers by year-end at least ensures a uniform baseline. But a mandatory tool whose detection logic is undocumented is a compliance posture, not a transparency guarantee. If Flock intends Audit Assistance to function as the accountability mechanism its announcements describe, the company will eventually need to subject that mechanism to outside scrutiny. An algorithm described only through anecdote is an algorithm whose limits nobody outside the vendor can assess.
