EU-Made Facial Recognition Ended up Scanning Schoolchildren in Brazil

Why It Matters

The lack of export controls lets EU‑origin AI tools operate in jurisdictions with weaker privacy safeguards, threatening vulnerable populations and undermining the EU’s credibility as a trustworthy AI benchmark.

Key Takeaways

• EU AI Act lacks export restrictions for biometric tools
• Innovatrics' system tracks nearly 1 million Brazilian students daily
• Accuracy falls short of contract 95% threshold, at 91%
• Legal challenge raises data‑protection and consent concerns in Brazil

Pulse Analysis

The European Union positioned its AI Act as a cornerstone for trustworthy artificial intelligence, imposing strict limits on biometric surveillance within its borders. While the legislation bans or heavily regulates such systems in schools and public spaces, it deliberately left export controls out of the final text. This omission creates a loophole where technologies deemed unacceptable at home can be sold abroad, raising questions about the EU’s commitment to its own ethical standards and the potential for regulatory arbitrage.

In Brazil’s Paraná state, Innovatrics—a Slovak company with €23 million in annual revenue—has rolled out a cloud‑based facial‑recognition attendance system across more than 1,700 schools, affecting nearly one million children each day. The algorithm, which matches classroom photos against a biometric database, records an average accuracy of 91.1%, below the 95% performance level stipulated in the procurement contract. Errors have practical consequences: attendance data feed into Bolsa Família eligibility, and teachers report slower processing and frequent misidentifications, prompting a legal challenge that cites Brazil’s data‑protection law and the lack of parental consent.

The Brazilian deployment highlights a broader international risk: without export oversight, EU‑origin AI can proliferate in markets with limited oversight, potentially eroding privacy rights and public trust. As the system expands to additional Brazilian states and even pilots in Portugal, regulators in both regions are beginning investigations, signaling growing scrutiny. Policymakers may need to revisit the AI Act to incorporate export‑control mechanisms, ensuring that the EU’s ethical AI framework extends beyond its borders and preserves its reputation as a global standard‑setter.