Brazil's ANPD gets 32% more incident reports while AI rules remain open
A review by Brazil's data protection authority shows 423 incident reports in the first nine months of 2026 and AI regulation still in progress. What your company should prepare now.

Brazil’s National Data Protection Authority (ANPD) received 423 security incident reports in the first nine months of 2026, about 32% more than the 321 in the same period of 2025. The total already exceeds the 395 for all of last year, according to a TeleSíntese report on the agency’s six-year review, the subject of an ANPD event today in Brasília.
How to read these numbers
The data measures what reached the agency, not the total number of incidents in the country. The report does not identify the cause of the increase, which could come from more attacks, more companies reporting, or both. The full report was not available in the sources consulted, so the numbers here come from TeleSíntese.
The review also shows what remains open. Of the 19 initiatives on the regulatory agenda, 16 are still in progress. Among them are the application of LGPD, Brazil’s data protection law, to AI, including Article 20, on automated decisions, minimum security standards and the use of biometric data.
The agency, on the other hand, has gained clout. Since February it has been a regulatory agency, and Capital Digital reports a fine of more than R$ 153 million against TikTok in August, for failures in protecting the data of children and teenagers.
What this changes for your company
1. Test your incident plan. Define who decides, who communicates and with what information. Inspections can look into unreported cases, identified through complaints, news or monitoring. Hiding the problem doesn’t end the risk.
2. Don’t wait for the AI rule to get organized. Article 20 of the LGPD already guarantees review of automated decisions. List where AI decides something about people, such as credit, hiring or customer service, and who is accountable for it. The free self-assessment helps find the gaps, and the ANPD sandbox showed that documents without evidence aren’t enough.
3. Document your use of biometrics. There is no final rule yet, but the ANPD already points to risks of error and discriminatory effects in facial recognition. Anyone using it at the front gate or in the time clock should record why and how failures are handled.
In 29 years of programming, I’ve learned that a good incident plan is one that was tested before it was needed.
Originally published on IWA7 Insights, IWA7 Digital’s blog, which is now part of Escovando Bytes.


