ANPD's AI sandbox has three companies testing explainability under the LGPD — and a second cycle announced for late 2026
Brazil's data protection authority opened a call in July 2025, selected Metatext, Synapse AI and Prevvine in October, and began operations in March 2026, with CIAAM/USP, to test transparency and explainability of automated decisions (LGPD art. 20). The first partial report came out in July 2026; the cycle runs to December. On 2 September, ANPD announced a second sandbox and an internal AI center.
Three participants; a first four-month cycle devoted to setup; first-sandbox results expected by the end of October 2026.
Tiny sample and narrow focus (art. 20). The sandbox does not create rules; it creates evidence for future rules. ANPD gained special-agency status in 2026 and is multiplying cases, but its technical capacity is still small.
It is where Brazil's rule on AI explainability is being written in practice — before the law.
Same theme
Regulation and Brazil→Get the next Cinco
Tuesday, 7am BRT, in your inbox. Five items, with sources. No daily newsletter, no promotions, and unsubscribing is one click.