Kemendagri Expands AI Bansos Verification Pilot to 258 Regions
Interior Minister Tito Karnavian said on September 30, 2026 that Indonesia is expanding an artificial-intelligence and population-data pilot that verifies and validates social-assistance (bansos) beneficiaries from an earlier 43 localities to 258 regencies and cities nationwide. Speaking after a ministerial meeting on social-protection interventions at the Social Affairs Ministry, Tito said the Banyuwangi first trial found roughly 60 to 70 percent mistargeting where ineligible people received aid while about 40 percent of eligible residents were still unregistered, prompting President Prabowo-era and National Economic Council guidance to scale the oversight programme.
Filed under Policy and dated October 2, 2026, this AI4Indonesia briefing treats the Perlinsos expansion as Indonesian public-sector AI targeting news distinct from yesterday’s West Java halt of the BDx Jatiluhur AI campus. The verification platform leans on about 98 percent biometric population coverage at Ditjen Dukcapil—face, fingerprint and iris—then integrates AI checks across eight ministries and agencies, including land ownership from ATR/BPN and inventory vehicle ownership from Polri, while voluntary Agen Perlinsos field agents help residents without smartphones.
Why it matters: Social budgets already leak when eligibility data lag reality. Cross-agency AI matching can tighten targeting—but only if appeal rights, biometric consent trails and human override stay explicit.
What it means in practice
Indonesian social-protection, privacy and counsel leads should inventory which bansos rolls will face AI asset-flagging next; demand named owners for Dukcapil biometric match accuracy and false-positive appeals; assign an owner for ATR/BPN and Polri data-sharing agreements; run time-boxed reviews of Agen Perlinsos field protocols for citizens without phones; and prefer designs that keep humans on final eligibility decisions. Place the pilot beside Komdigi’s presidential AI roadmap and ethics drafts and West Java’s BDx Jatiluhur halt.
Caveats come first. A 258-region pilot is not finished national coverage; asset flags can misread shared family property; and biometric matching raises exclusion risks. AI4Indonesia therefore presents the expansion as directional policy context until published precision metrics appear.
What to watch next: published mistargeting rates after the wider wave; complaint channels for wrongly flagged households; and how Komdigi’s ethics Perpres treats social-protection AI. Readers can continue on the AI4Indonesia homepage, or browse the Newsroom for additional briefings.
Bottom line: treat this update as orientation, not instruction. Indonesian social AI is pairing biometrics with multi-agency asset checks and remains early. Organizations that benefit most will demand appeal evidence, keep humans on eligibility gates, and refuse to confuse a pilot expansion with finished fair targeting.