- Why in news
- On 20 February 2026, the Ministry of Electronics & IT said UIDAI has deployed an AI-enabled biometric deduplication and document verification platform for Aadhaar enrolment and updates.
- What it is
- Aadhaar is a 12-digit unique identity number issued by UIDAI to residents of India. Deduplication checks each new enrolment's biometrics (fingerprint, face, iris) against the existing database so that one person gets only one number. The new AI layer aims to make these checks more accurate and to verify documents faster.
- Biometric modalities covered
- AI models have been built for fingerprint, face and iris matching.
- Development partner
- The models were developed with IIIT Hyderabad.
- Computing infrastructure
- The models run on NVIDIA DGX high-performance inference infrastructure.
- Scope of check
- Every new enrolment is cross-checked against the whole Aadhaar database to ensure uniqueness.
- Document verification
- AI extracts document metadata and verifies it against the source of truth using DigiLocker APIs.
- Rollout status
- Enhanced deduplication is done for several states, with nationwide completion expected in the coming months.
Don’t confuse with Deduplication vs authentication: Deduplication happens at enrolment or update to ensure one person has one number. Authentication later checks that a person matches their existing Aadhaar.
Don’t confuse with UIDAI vs DigiLocker: UIDAI issues Aadhaar. DigiLocker is a separate digital document platform whose APIs UIDAI uses to verify documents.
Science and technology: 57 Prelims questions since 2013; most test the concept behind the news, not the news itself. In Mains, 12 news-triggered questions.
Why it matters
- Prevents duplicate or fake identities, protecting welfare delivery and the integrity of the identity database.
- Faster, more accurate document checks can cut enrolment and update turnaround time.
Problems
- Biometric matching can produce false rejections, for example for manual labourers and the elderly with worn fingerprints.
- Large centralised biometric data raises privacy and data-protection concerns, and dependence on foreign hardware raises questions of self-reliance.
Way forward
- Provide non-biometric fallbacks so no resident is denied services due to a failed match.
- Run audits and apply data-protection safeguards under the Digital Personal Data Protection Act, 2023, and expand indigenous computing capacity.
Put in your answer
Indigenous AI model example: UIDAI built AI models for fingerprint, face and iris with IIIT Hyderabad, an example of an institutional partnership for digital public infrastructure.
- 2014Mains GS2 · Q15Whether Aadhaar and NPR should run concurrently, and their developmental potential.
- 2026Mains GS4 · Q8AI biometric crowd-matching in predictive policing; ethical dilemmas for the officer.
- 2014Prelims Q66Which methods besides fingerprint scanning can biometrically identify a person?
- 2018Prelims Q17What open Aadhaar APIs imply: device integration and iris-based online authentication.
UPSC’s own questions that come close to this issue, from our Prelims and Mains banks.
No practice question for this one; the card may still help an answer. Why: Lasting topic on Aadhaar and AI use in governance, but few testable facts and no clear debate for a question.
Worth revising until Prelims 2027: most news questions come within two years of the event.