A Regional Watchlist Gets an AI Upgrade
At the close of the 44th ASEANAPOL Conference, hosted by the Philippine National Police in Manila, member states endorsed a long-term modernization plan for the electronic ASEANAPOL Database System (e-ADS) — the regional repository used by all eleven ASEAN police forces to track suspects and fugitives across borders. PNP Chief Gen. Jose Melencio Nartatez Jr. directed Philippine units to maximize facial recognition tools for suspects with "potential international links," telling reporters that "the responsible use of facial recognition technology will enable faster verification of identities, more efficient exchange of intelligence, and quicker coordination among Asean police organizations."
ASEANAPOL, the 1981-founded association of Southeast Asian police forces, describes the plan as a shift "from e-ADS 2.0 toward an artificial intelligence-enabled system" built for advanced analytics and cross-border correlation. Timor-Leste was admitted as the bloc's eleventh member at the same conference, expanding the database's jurisdictional footprint just as its capabilities grow.
The Case For It
Start with the strongest version of the argument for this. ASEAN's internal borders are porous, its member states have wildly uneven digital forensics capacity, and fugitives routinely exploit that gap — flee a warrant in one jurisdiction, resurface under a new identity two countries away. Manual name-and-photo matching across eleven separate national databases is slow enough that it functions as a de facto amnesty for anyone with a plane ticket. A shared, faster identity-matching layer targeting people already wanted on outstanding warrants is a legitimate, proportionate use of the technology — closer to Interpol's Red Notice system than to mass surveillance of the public. The conference's other priorities, expanded forensic cooperation and a joint task force against cybercrime, scams, and trafficking through 2028, point at real transnational harms, not manufactured ones.
What Manila Hasn't Said
The problem is what's missing from the announcement itself. As industry coverage of the rollout noted, the PNP's statement did not name a facial recognition vendor, specify an accuracy threshold, disclose an image retention policy, or explain a budget or deployment timeline. The PNP says any match would trigger "human review and confirmation," but offered no detail on how a wrongly flagged person could challenge a match, or which of the eleven member states' records that person's face would already have touched by the time the error was caught. Those gaps matter because facial recognition accuracy is not a fixed property of "the technology" — it varies enormously by which algorithm is deployed and under what conditions.
NIST's Face Recognition Vendor Test, the most rigorous independent benchmark of its kind, found in testimony to Congress that top-performing algorithms failed in roughly one in 400 searches under NIST's controlled 2018 mugshot test — but also that false-positive rates "often vary by factors of 10 to beyond 100 times" across demographic groups, with women, younger and older adults, and Asian and African American faces bearing the worst of it. A system built on an unvetted or unbenchmarked vendor could easily land far outside NIST's best-case numbers. ASEANAPOL's masterplan, as described, commits to expanding matching capacity without committing to publish which algorithm clears which bar.
The US Is Fighting Over the Same Question
Washington is having a version of this argument in public, which is instructive by contrast. Sens. Ron Wyden and Jeff Merkley, joined by Sen. Ed Markey and Rep. Pramila Jayapal, introduced the ICE Out of Our Faces Act on February 6, 2026, to bar ICE and CBP from using facial recognition and to force deletion of biometric data already collected, after reports that agents deployed the tool on people in public without consent. A separate bill, the DHS Surveillance Technology Moratorium Act, would pause covered DHS surveillance contracts pending an Inspector General audit of what's actually in use. Both bills are contested — DHS argues restrictions would hamper legitimate enforcement — but the fight itself forces a public accounting of vendor, accuracy, and retention questions before deployment scales further. ASEANAPOL's announcement skipped that step entirely; there is no equivalent floor debate, no bill text, no named vendor to scrutinize.
Attach the Conditions Before the System Goes Live
This publication doesn't oppose facial recognition for cross-border fugitive identification — that's a narrower, more defensible use case than domestic street surveillance, and the underlying crime problem ASEANAPOL cites is real. But proportionate regulation means the accuracy and accountability rules get built into the procurement decision, not bolted on after a wrongful match makes headlines. ASEANAPOL's modernization masterplan is the moment to require independent, NIST-style benchmark testing before any vendor is approved, a published retention and deletion schedule, and a defined redress channel that works across all eleven member states' legal systems — not just the Philippines'. Regional biometric sharing is being built now, while conditions are still cheap to attach. Retrofitting them after e-ADS 3.0 is live, across a system where a bad match in one country propagates to ten others, will not be.