BRA-002

INSS Meu INSS AI-Enabled Benefits Adjudication System

Download PDF
Brazil Latin America & Caribbean Upper middle income Full Production Deployment Confirmed

Instituto Nacional do Seguro Social (INSS); Dataprev (state-owned technology company, system developer)

At a Glance

What it does Perception and extraction from unstructured inputs — Decision support for eligibility and benefits
Who runs it Instituto Nacional do Seguro Social (INSS); Dataprev (state-owned technology company, system developer)
Programme INSS Social Security Benefits (Retirement, Sick Pay, Disability) via Meu INSS Application
Confidence Confirmed
Deployment Status Full Production Deployment
Key Risks Model-related risks
Key Outcomes INSS states that automated decisions are based on specified legal criteria.
Source Quality 3 sources — Government website / press release, News article / media, Academic journal article

Brazil's Instituto Nacional do Seguro Social (INSS), through the state-owned technology company Dataprev, deployed an AI-enabled claims adjudication engine within the Meu INSS mobile application beginning in 2018. The system uses computer vision and natural language processing to scan data in documents uploaded by social security claimants and automatically processes welfare claims including retirement, sick pay, and disability benefits. This is distinct from the INSS Helo chatbot (BRA-001), which provides informational assistance; the adjudication engine makes actual approval or denial decisions on benefit claims.

The system was introduced to address a backlog of approximately 2 million pending benefit requests. The Brazilian government has set a target for the algorithm to review 55 percent of all social security filings by the end of 2025. The Meu INSS app receives nearly 84 million hits per month, making it one of the largest AI-assisted benefits adjudication systems in Latin America.

A detailed investigation by Rest of World (April 2025) documented significant problems with the system's automated decision-making. Joselia de Brito, a 55-year-old former sugarcane worker from northeast Brazil with chronic conditions including herniated disc, scoliosis, and fibromyalgia, had her retirement claim rejected in February 2025 because the system misidentified her as male. Her claim was only approved in March after intervention by the National Confederation of Workers in Agriculture.

The investigation identified several systematic failures: the system issues automatic rejections for minor errors in applications; it cannot properly analyse complex filings from agricultural workers whose documentation differs from standard employment records; it inadequately handles cases involving hazardous working conditions which require additional paperwork; and it has difficulty processing cases involving shared land ownership. Approximately 34,300 benefits were denied for rural workers in January 2025 alone.

Access to appeal is also problematic. The average wait for the internal legal resource board to review an appeal is 278 days. Jane Berwanger, Director of the Brazilian Institute for Social Security Law, warned that unreviewed automated claims 'will turn into legal battles.'

The system disproportionately affects rural populations. Brazil's rural illiteracy rate was 15 percent in 2022, three times higher than urban areas. Some applicants must travel four hours to the nearest INSS office. Edjane Rodrigues, Secretary for social policies at the National Confederation of Workers in Agriculture, noted that automatic denials with little recourse particularly harm these vulnerable populations.

An INSS spokesperson stated that 'each automated decision is based on specified legal criteria, ensuring that the standards set by the social security legislation are respected.'

In January 2025, Dataprev announced approximately $10.5 million in investment to enhance data analysis and fraud detection. In February 2025, a new AI feature was introduced for personalised user offerings.

The system is internally known as 'Isaac' (named after Isaac Asimov), as documented in Dataprev's official 2019 announcement at their 5th Innovation Week. Dataprev described Isaac as using predictive algorithms and machine learning to cross-reference multiple databases for automatic benefit concession, with the ability to process thousands of applications in parallel using adjustable risk criteria. The system also integrates biometric facial and fingerprint verification. At the time of launch, approximately 40 percent of benefit requests faced denial with an 89-day average analysis period.

A peer-reviewed study published in Internet Policy Review (Nicolas and Sampaio, 2024) provided the most comprehensive independent analysis of the system's impacts. The study documented that in 2022 INSS automatically rejected more than 800,000 applications, representing a 300 percent increase over 2021. Urban maternity benefit rejections surged from 7,064 to 60,379, with 85.9 percent being automatic decisions. The automatic analysis rate rose from 17 percent in 2022 to 36 percent in 2023. The study also found that the CNIS database (Cadastro Nacional de Informacoes Sociais), which underpins the automated system, had 24.3 million entries with incomplete or invalid data, raising fundamental questions about the reliability of automated decisions built on this foundation.

Brazil's Federal Court of Accounts (Tribunal de Contas da Uniao, TCU) conducted an audit in 2024 that found significant data quality issues in CNIS affecting automatic benefit concession, including incomplete employment data, inconsistent records, employment start dates predating company registration, and records linked to non-existent company registration numbers. The TCU imposed a 180-day deadline on INSS and Dataprev to address these issues.

Classifications follow the DCI AI Hub Taxonomy. Hover over field labels for definitions.

Social Protection Functions

Implementation/delivery chain
Assessment of needs/conditions + enrolment primaryProvision of payments/services
SP Pillar (Primary) The social protection branch: social assistance, social insurance, or labour market programmes. Social insurance
SP Pillar (Secondary) The social protection branch: social assistance, social insurance, or labour market programmes. Social assistance
Programme Name INSS Social Security Benefits (Retirement, Sick Pay, Disability) via Meu INSS Application
Programme Type The type of social protection programme, classified under social assistance, social insurance, or labour market programmes. View in glossary Old age, survivors and disability pensions
System Level Where in the social protection system the AI is applied: policy level, programme design, or implementation/delivery chain. View in glossary Implementation/delivery chain
Automation Subtype For operational automation cases: (a) document processing and generative staff assistance, or (b) workload and resource forecasting. (a) Document processing and generative staff assistance
Programme Description AI-enabled automated benefits adjudication engine within Brazil's Meu INSS mobile application that uses computer vision and NLP to scan uploaded documents and process social security claims including retirement, sick pay, and disability benefits.
Implementation Type How the AI output is produced: Classical ML, Deep learning, Foundation model, or Hybrid. Affects validation, compute requirements, and governance profile. View in glossary Deep learning
Lifecycle Stage Current stage in the AI lifecycle, from problem identification through to monitoring, maintenance and decommissioning. View in glossary Integration and Deployment
Model Provenance Origin of the AI model: developed in-house, adapted from open-source, commercial/proprietary, or accessed via third-party API. View in glossary Developed in-house
Compute Environment Where the AI system runs: on-premise, government cloud, commercial cloud, or edge/device. View in glossary Not documented
Sovereignty Quadrant Classification of data and compute sovereignty: I (Sovereign), II (Federated/Hybrid), III (Cloud with safeguards), or IV (Shared Innovation Zone). View in glossary Not assessed
Data Residency Where the data used by the AI system is stored: domestic, regional, or international. View in glossary Not documented
Cross-Border Transfer Whether data crosses national borders, and if so, whether documented safeguards are in place. View in glossary Not documented
Is Agentic Whether the system autonomously plans and executes multi-step workflows, selecting tools and chaining actions with limited human intervention. View in glossary Partial
Agentic Pipeline Description of the chained workflow steps in the agentic pipeline. Document scanning (CV/NLP) -> eligibility criteria matching -> automated approval/denial decision
Agentic Autonomy Degree of autonomy: fully autonomous, semi-autonomous (human checkpoints), or supervised (human approval at each step). Semi-autonomous
Override Points Where in the pipeline human review or override is triggered. Internal legal resource board appeal (average 278-day wait); intervention by advocacy organisations
Decision Criticality The rights impact of the decision the AI supports. High criticality requires HITL oversight; moderate requires HOTL; low may operate HOOTL. View in glossary High
Human Oversight Type Level of human involvement: Human-in-the-Loop (active review), Human-on-the-Loop (monitoring), or Human-out-of-the-Loop (periodic audit). View in glossary HOTL
Development Process Whether the AI system was developed fully in-house, through a mix of in-house and third-party, or fully by an external provider. View in glossary Fully in-house
Highest Risk Category The most significant structural risk source identified: data, model, operational, governance, or market/sovereignty risks. View in glossary Model-related risks
Risk Assessment Status Whether a formal risk assessment, informal assessment, or independent audit has been conducted for this system. Not assessed
Documented Risk Events System misidentified a female claimant as male, causing wrongful rejection. Automatic rejections for minor application errors. Inability to process complex agricultural worker filings. Inadequate handling of hazardous working conditions documentation. Difficulty with shared land ownership cases. 278-day average appeal wait time.

Risk Dimensions

Data-related risks
Governance and institutional oversight risks
Operational and system integration risks

Impact Dimensions

Autonomy, human dignity and due process
Systemic and societal
  • Grievance mechanism
CategorySensitivityCross-System LinkageAvailabilityKey Constraints
Beneficiary registries and MISSensitiveSingle source (no linkage)Currently available and usedDocuments uploaded by claimants through the Meu INSS app including identification documents, employment records, medical certificates, land ownership records, and hazardous conditions documentation. Computer vision and NLP used to extract data from scanned documents.

Dataprev (2019) '5a Semana de Inovacao: Dataprev apresenta Isaac, solucao de IA', Portal Dataprev, 7 November. Available at: https://portal.dataprev.gov.br/5a-semana-de-inovacao-dataprev-apresenta-isaac-solucao-de-ia (Accessed: 30 March 2026).

View source Government website / press release

Daros, G. (2025) 'Brazil's AI-powered social security app is wrongly rejecting claims', Rest of World, 24 April. Available at: https://restofworld.org/2025/brazil-ai-social-security-app-rejected/ (Accessed: 30 March 2026).

View source News article / media

Nicolas, M.A. and Sampaio, R.C. (2024) 'Balancing efficiency and public interest: The impact of AI automation on social benefit provision in Brazil', Internet Policy Review, 13(3). doi: 10.14763/2024.3.1799.

View source Academic journal article
Deployment Status How far the system has progressed into real-world operational use, from concept/exploration through to scaled and institutionalised. View in glossary Full Production Deployment
Year Initiated The year the AI system was first initiated or development began. 2018
Scale / Coverage The scale and geographic or population coverage of the deployment. 84 million app hits per month; government target of 55% of all social security filings reviewed by algorithm by end of 2025; approximately 34,300 rural worker claims denied in January 2025
Funding Source The source(s) of funding for the AI system development and deployment. Brazilian federal government budget; Dataprev investment of approximately $10.5 million (January 2025) for enhanced data analysis and fraud detection
Technical Partners External technology vendors, academic partners, or development partners involved. Dataprev (state-owned technology company)
Outcomes / Results INSS states that automated decisions are based on specified legal criteria. System was introduced to address a backlog of approximately 2 million pending requests. Approximately 34,300 rural worker benefits denied in January 2025 (down from 53,400 one year prior). Documented cases of wrongful denials requiring external intervention to overturn.
Challenges Disproportionate impact on rural populations with 15% illiteracy rate. Nearest INSS office four hours away for some applicants. System cannot handle complex or non-standard documentation common among agricultural workers. Appeal process averaging 278 days creates effective denial of recourse. Digital literacy barriers for vulnerable populations.

How to Cite

DCI AI Hub (2026). 'INSS Meu INSS AI-Enabled Benefits Adjudication System', AI Hub AI Tracker, case BRA-002. Digital Convergence Initiative. Available at: https://socialprotectionai.org/use-case/BRA-002 [Accessed: 1 April 2026].

Change History

Updated 1 Apr 2026, 08:11
by system (system)
Updated 31 Mar 2026, 06:35
by system (system)
Created 30 Mar 2026, 11:18
by v2-import (import)