Brazil's Social Security AI Reportedly Rejecting Complex Claims Improperly
Brazil's social security AI system Meu INSS reportedly rejects retirement and disability claims from rural workers due to system errors without effective review mechanisms. The AI misidentified claimant Josélia de Brito as male and rejected her application; experts report the system disproportionately harms those with limited technology access.
What happened
The Meu INSS application, a social security system deployed by Brazil's Instituto Nacional do Seguro Social (INSS) in 2018, recently incorporated artificial intelligence to automate benefit claim processing. The government aimed for 55% of claims to be analyzed automatically by 2025.
Josélia de Brito, an ex-rural worker from northeastern Brazil, had her retirement claim automatically rejected due to a system error that misidentified her as male. According to reports, the system rejected claims with small errors or inconsistencies without offering effective review alternatives.
Reported impact
Rural workers and people with limited technology access have faced major difficulties accessing Meu INSS services. Rural areas have literacy rates three times higher than urban zones and lack reliable internet, making interaction with the digital system more challenging.
Josélia de Brito reported frustration at repeated claim rejections despite having required documentation, describing the process as humiliating when system errors resulted in denials without clear explanations.
Index analysis
Experts alert that automation failures can leave the most vulnerable without support, creating environments where efficiency does not translate to fairness. The system lacks adequate human review mechanisms for complex cases, particularly affecting those in remote areas.
Governance controls
The system does not offer effective alternatives for reviewing or correcting inconsistencies identified by the AI. Specialists note that technology implementation without critical analysis and adequate citizen training has resulted in digital exclusion.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.