ANPD Sandbox Report Details AI and Data Protection Tests by Metatext, Synapse AI and Prevvine
Brazil’s National Data Protection Authority (ANPD) has published the second partial monitoring report from its regulatory sandbox on artificial intelligence and personal data, IBTimes Brasil reported on October 12. The document, released on October 9, sets out progress and problems seen in tests by three companies in a pilot supervised by the agency together with the Center for Artificial Intelligence and Machine Learning at the University of São Paulo (Ciaam/USP).
The participants are Metatext, Synapse AI and Prevvine Tecnologia. They entered the experimental environment in March 2026 and are building tools that use AI to process personal data in finance, mobility and health. According to the ANPD, the monitoring stage included training on cybersecurity, risk management and regulatory compliance. The companies put their systems through security and compliance tests, and critical points that surfaced went through rounds of improvement and fresh assessment. Criteria were adjusted to the technological maturity of each project rather than applying identical technical checks to systems with different purposes.
A central goal is to see how the tools treat personal data and whether they respect the guarantees of the General Data Protection Law (LGPD). Transparency, explainability, security and the ability to identify risks matter most where automated decisions can affect consumers, patients or other citizens. The report is not a blanket approval: taking part means being assessed under supervision, and it does not remove the responsibility of the companies for how they handle data.
What it means in practice
For companies building AI on personal data, the sandbox is an early signal of what the regulator will look for, from documented risk management to explanations people can understand. The ANPD says what it learns could feed future guidance and rules on responsible AI use, aiming to reconcile innovation with fundamental rights and legal certainty. The project has also been named a finalist in the Global Privacy and Data Protection Awards 2026 from the Global Privacy Assembly.
The report follows plans covered in the internal AI centre and second sandbox at the ANPD, and arrives while the AI framework vote has slipped past the election calendar. What to watch: the final sandbox report, which lessons become formal guidance, and how the second sandbox selects its participants.