This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
The Pakistani National Database and Registration Authority (NADRA) implemented a biometric identification system that, while intended to improve service delivery and security, has caused widespread harm. Automated rules within the system have led to the mass suspension of hundreds of thousands of national identity cards (CNICs), effectively rendering citizens stateless and denying them access to basic rights, welfare, and financial services. Courts have repeatedly ruled that NADRA's automated adjudication of family and paternity disputes is illegal and lacks jurisdiction.
When the leader of the Afghan Taliban was found possessing a valid ID card in the Pakistani national biometric identification database system, Pakistan launch a national re-verification campaign that is linked to numerous changes in recognition status and loss of services.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that fail to perform reliably or effectively under varying conditions, exposing them to errors and failures that can have significant consequences, especially in critical applications or areas that require moral reasoning.
Also classified as
The biometric system performed unequally, failing to authenticate manual laborers with worn fingerprints and disproportionately impacting the Pashtun community.
The database suffered massive leaks exposing 100 million citizens' data, and employees abused their access to retrieve private phone numbers.
The system's rigid automated rules and biometric matching failures (such as facial recognition errors and cascading family tree blocks) demonstrate a lack of capability to robustly handle real-world complexities.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The widespread harms resulted from the post-deployment execution of automated database rules and biometric matching algorithms, which produced unexpected negative outcomes.
Added categories peak at 3: Substantial (Psychological Harm, Inferred)
The automated blockings and data leaks within Pakistans biometric database system (NADRA) led to severe civil rights violations, rendering over 600,000 citizens stateless and exposing personal data of 100 million people. This highlights the severe national security and societal stability risks of unregulated state-level algorithmic identity management.
The NADRA biometric database operates as a tool that performs specific, well-defined functions such as fingerprint matching, facial recognition, and executing pre-defined database rules based on direct inputs.
Risk Level 2: High Risk. The system involves biometric identification and facial recognition used by government authorities, which has significant implications for fundamental rights and access to essential public services.
The system's primary purpose is the biometric identification and registration of citizens, which is then used to allocate state resources and welfare benefits.
The reports provide extremely detailed, multi-perspective accounts of NADRA's operations, including specific court cases, personal testimonies, and historical context from the former chairman himself.
The issues could be viewed purely as administrative or political governance failures rather than AI safety failures, but the central role of automated database rules and biometric mismatches justifies the AI safety framing.
Technology
Flawed Biometric Matching Algorithms
Algorithms fail on worn fingerprints or produce facial recognition mismatches.
Inflexible Family Tree Rules
Hardcoded database rules block citizens based on rigid family structures.
Vulnerable Database Security
Lack of access controls allows data leaks and employee misuse of personal data.
Data Inputs
Unverified Family Linkage Data
Single flagged ID cascades blocks to all linked family members in the system.
Incomplete Historical Records
Demanding pre-1978 residency proof excludes citizens lacking physical papers.
Human Factors
Overreliance on Automated Logic
Officials trust system flags over obvious physical evidence of separate people.
Employee Misconduct
Staff abuse database access to retrieve contact details for personal harassment.
Lack of Digital Literacy
Illiterate users share PINs, leading to identity theft and fraud.
Process and Methods
Arbitrary CNIC Blocking Process
System suspends identities without prior notice, verification, or due process.
Inadequate Redress Mechanisms
Citizens face immense bureaucratic hurdles to resolve false system flags.
Management
Prioritizing Security over Safety
Rapid digitization prioritized state surveillance over citizen welfare safeguards.
Defensive Organizational Culture
Management refuses to acknowledge flaws in their automated matching systems.
Unauthorized Power Adjudication
NADRA management unilaterally resolves complex family and paternity disputes.
Regulatory Environment
Absence of Data Protection Law
No legal framework exists to protect citizen data privacy or hold NADRA liable.
Judicial and State Overreach
Courts and agencies use identity blocking as a tool for political punishment.
Lack of Independent Oversight
No external regulatory body monitors NADRA's automated decision-making.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
