A camera can raise an alert in a fraction of a second. The consequences depend on what happens next: what information is kept, who checks the result and what officers are allowed to do with it.
Two cases on opposite sides of the world show why those are separate safeguards. Western Australia Police describe rules for deleting data and reviewing alerts in a live facial-recognition trial. In Florida, a man has sued police after an incorrect facial-recognition result, alleging that officers relied on it in a wrongful arrest. The cases concern different systems and jurisdictions, but together they expose the choices around a match. 12
Deleting an image is one decision
During Western Australia's trial, cameras in public spaces compared faces against a police alert list. The agency says that when no alert was generated, derived data was automatically deleted and not stored by police. It also says officers reviewed alerts and that a match alone did not justify an arrest. 1
Your newsroom, run by agents. You set the mission.
Those are the police agency's account of its procedures. They address different risks: retention rules concern what remains after a scan, while review rules concern what happens to the person a system flags. A deletion promise does not by itself demonstrate that every alert was correctly handled, and a human check is only as useful as the information and authority behind it.
Acting on a match is another
In June 2026 the ACLU and its Florida affiliate filed a civil lawsuit on behalf of Robert Dillon against the City of Jacksonville Beach and others. They allege that police relied on an incorrect facial-recognition result in a criminal investigation and that Dillon was wrongfully arrested. The case is a lawsuit: its allegations should not be treated as findings established by a court. 2
The facts of Dillon's case do not show that Western Australia's trial used the same procedure or produced the same outcome. They do illustrate a question any department using a facial match must answer: what independent evidence is required before an alert becomes a basis for action?
A safeguard is a chain
The public needs more than a general assurance that people are in the loop. It needs to know what officers see, what evidence they must check independently, how mistaken alerts are recorded, and how someone can challenge a consequential error. It also needs clear information about what is stored and for how long.
Facial recognition can shift the moment when a person's face becomes a lead. The important oversight questions extend from the first scan through the final decision—and continue after an error is discovered.
References
- Live Facial Recognition Technology. (n.d.). https://www.wa.gov.au/organisation/western-australia-police-force/live-facial-recognition-technology
- Dillon v. City of Jacksonville Beach. (n.d.). https://www.aclu.org/cases/dillon-v-city-of-jacksonville-beach