AI Street-Homelessness Detection — Uruguay Repurposes Police Camera Analytics for Welfare Outreach, and Cannot Yet Measure It

In August 2023, Uruguay's Ministerio del Interior announced it would extend the AI behavioural-pattern recognition built for its Unified Command Centre (CCU) camera network - originally designed to flag suspicious criminal movement - to a new purpose: identifying people sleeping or lingering for extended periods on city streets.

The stated aim was to use existing surveillance infrastructure to detect street homelessness and trigger a welfare response, rather than a purely security one.

When the system flags a likely case, CCU video operators verify it visually and dispatch a joint response with the Ministerio de Desarrollo Social (Mides), whose street teams then try to persuade the person to accept a shelter place. The tool draws on the same camera infrastructure used for crime prevention - up to roughly 11,900-12,000 cameras nationally - and was reported operational by the end of 2023, remaining active through Uruguay's June 2025 cold-wave emergency.

No source we found publishes a detection count, a confirmation rate, or any measure of how many additional people were sheltered because of the AI alert versus routine patrols. Officials have only said 'the number of homeless people assisted grows daily,' without a figure, and Mides's own leadership acknowledges that how many go undetected is 'difficult to gauge.' Montevideo's 2023 street-count census separately recorded 2,755 people sleeping rough, a 24% rise since 2021, with 91% reporting substance-use issues.

Using AI-linked policing infrastructure to surveil an already-vulnerable population, with no public accuracy or outcome data, is a real governance and evidence gap that this entry records honestly rather than smooths over.