The government of Michoacán, Mexico has launched a pilot of live facial recognition using the SAFR platform to help identify people with outstanding arrest warrants and locate missing persons. The program links surveillance cameras across urban centers, rural areas, and state roads with the State Center for Command, Communications, Computing, Control, Coordination and Intelligence (C5), which routes alerts to relevant authorities for response.
How the Pilot Works
The system analyzes live video feeds and detects faces that may match records in reference databases managed by the federal Office of the Attorney General and the state prosecutor. According to local officials, integrations also extend to certain private security systems operated by banks and Oxxo convenience stores through the C5. When the software finds a possible match, it generates an immediate alert for law enforcement personnel, allowing officers to verify and act. Authorities say the pilot database currently contains 319 individuals, including missing persons, and they claim the system reaches 99.87 percent accuracy. The government has not published technical details on the tests behind the accuracy figure, the false match rate, or how performance will be monitored during live deployment.
The SAFR platform is recognized for operating on live video streams and for supporting real-time alerting workflows. In Michoacán, the state is using that capability to prioritize alerts for wanted suspects and persons reported missing. Officials say the goal is to speed investigative leads and improve situational awareness by connecting camera networks and relevant records under a single operational flow at the C5.
Privacy, Safeguards, and Open Questions
Human rights and privacy advocates in Mexico have warned that broad use of public camera networks for biometric identification can normalize mass surveillance and increase the risk of misidentification. They have also called for clear limits on data retention and access. Michoacán officials say the pilot follows security protocols and is restricted to law enforcement purposes, but public documentation to date provides limited detail on governance, audit processes, or how long facial images and alerts will be stored. It is also unclear what recourse exists for individuals who believe they were wrongly flagged, and whether independent accuracy testing or bias evaluation will be conducted.
Next Steps
State authorities indicate that results from the pilot will guide decisions on expanding coverage to additional locations and use cases. Key metrics are expected to include the number of validated matches, response times, and outcomes in missing person and warrant cases. Observers in the public safety and biometrics communities continue to track the initiative to assess effectiveness, proportionality, and adherence to privacy and human rights standards.
Source: OECD Policy Observatory
–
By the ID Tech Editorial Team
