Brazil
AI behavior law: bot and agent disclosure, crawler and training-data rules, automated-agent transactions, and algorithmic decision-making.
Summary
Brazil presents moderate risk. The 'Lei Carolina Dieckmann' (Law 12.737/2012, inserting Penal Code Art. 154-A) criminalizes 'improper violation of a security mechanism' to access computer data; public pages generally fall below this threshold. The Copyright Law 9.610/1998 has a closed exception list with no TDM provision, making commercial AI training likely prohibited under copyright. Marco Civil da Internet (Law 12.965/2014) governs internet use and data protection obligations. LGPD (Law 13.709/2018) applies robustly: the ANPD in 2024 issued a preventive measure halting Meta from training AI on Brazilian public social-media data, signaling aggressive enforcement. PL 2338/2023 (AI bill) includes proposed rightholder opt-out requirements and is pending.
Automated-access legality
Carried forward from the crawler-law index. Governs whether automated clients may access public websites in this jurisdiction.
| Dimension | Value |
|---|---|
| Authorization test | security mechanism bypass |
| Public-page carve-out | yes |
| Terms-of-service browsewrap enforceable | notice dependent |
| Terms-of-service clickwrap enforceable | yes |
| Copyright exception model | closed list |
| Text and data mining — commercial status | prohibited |
| Text and data mining — opt-out mechanism | none |
| robots.txt legal weight | non binding notice |
| AI training-specific law | pending |
| Privacy regime | LGPD |
| Trespass to chattels | not recognized |
Last reviewed: 2026-05-21. Confidence: medium. Fast-moving area — verify before relying. Not legal advice.