Summary
- Risk tiers (Senate text): "excessive risk" uses are banned (Art. 13), such as manipulation causing harm, crime-risk scoring based on personality or past behaviour, and AI that produces child sexual abuse material. "High risk" uses are listed in Art. 14, including critical infrastructure, student selection, hiring, credit-like access to essential services, and emotion recognition.
- Governance duties for high-risk systems (Arts. 18 to 24), an algorithmic impact assessment before market launch (Arts. 25 to 28, conclusions made public), and specific duties for general-purpose and generative AI developers (Arts. 29 to 33). Synthetic content must carry identifiers (Art. 19).
- Rights for people affected by high-risk AI, including explanation and human supervision (Arts. 6 to 9).
- Governance system (SIA): a National AI Regulation and Governance System coordinated by the ANPD, working with sectoral regulators (Arts. 45 to 49).
- Copyright: developers using protected works must disclose this and rights holders can opt out and seek remuneration (Arts. 62 to 65).
- Status: proposed. As of the Chamber's open-data record (latest entry 2026-09-02) the bill sits in the Special Committee "awaiting report". No committee or plenary vote found.
Who it applies to
- Developers, distributors and deployers ("agentes de IA") of AI systems in Brazil (Art. 1).
- Excluded: purely personal non-economic use, national defence only, pre-market research and testing, and pure storage or data transport infrastructure (Art. 1, §1).
- No explicit extraterritorial clause found in the Senate text. Our reading: a foreign provider whose AI is offered or used in Brazil would still be in scope as an AI agent operating in Brazil.
Key dates
- Bill presented in the Federal Senate (Senator Rodrigo Pacheco)
- Senate plenary approves substitute text (rapporteur Senator Eduardo Gomes)
- Received and presented in the Chamber of Deputies
- Chamber constitutes a Special Committee to report on the bill
- Deputy Aguinaldo Ribeiro (PP-PB) named rapporteur
- Latest Chamber record: status 'Aguardando Parecer' (awaiting rapporteur's report), priority regime
Full date record
| Date | Event |
|---|---|
| 2023-05-03 | Presented in the Senate |
| 2024-12-10 | Senate plenary approves substitute |
| 2025-03-17 | Received by the Chamber of Deputies |
| 2025-04-29 | Special Committee constituted |
| 2025-05-20 | Rapporteur named (Aguinaldo Ribeiro) |
| 2026-09-02 | Latest record: awaiting report; many related bills attached |
| If enacted | Most rules apply 730 days after publication; bans (Art. 13) and generative AI rules after 180 days; governance chapter on publication (Art. 80, Senate text) |
Penalties and enforcement
- Senate text Art. 50: warning; fine up to R$ 50 million per infraction, or up to 2% of the group's Brazilian gross revenue for private companies; publication of the infraction; bans from the sandbox up to 5 years; suspension of development or use.
- Enforcement would sit with the ANPD and sectoral regulators. None of this applies until enacted.
- Text can still change substantially in the Chamber. If the Chamber amends it, it returns to the Senate.
What it means for an Indian company
- No legal duty today. Watch it if you sell AI products in Brazil, especially HR tech, credit or fintech, edtech or biometric tools that fall into Art. 14.
- Prepare for an EU-style regime: a risk classification of each use case (SC-AI-004 AI risk classification and use-case tiering), a pre-launch impact assessment that may need to be published (SC-AI-006 AI impact assessment (people and society)), and explanation plus human review rights (SC-AI-027 Right to contest and human review for individuals).
- Generative AI vendors: plan for synthetic content labelling (SC-AI-023 Label AI-generated and synthetic content) and training data copyright records (SC-AI-010 Training and input data governance (provenance, quality)).
- Brazil's data law already applies now: see Brazil LGPD.
Obligations
Full obligation breakdown coming
We break the most requested policies down into numbered obligations mapped to controls. Leave your details and we’ll tell you when Brazil AI Bill (PL 2338/2023) is done.
Glossary
- PL: Projeto de Lei, a bill.
- SIA: Sistema Nacional de Regulação e Governança de Inteligência Artificial, the proposed AI governance network.
- ANPD: Brazil's National Data Protection Agency, proposed as SIA coordinator.
- Comissão Especial: a special committee created to report on a bill that touches many subject committees.
- Parecer: the rapporteur's report and recommended text.
- Excessive risk: the bill's term for banned AI uses.
Key takeaways
- Not law. Passed the Senate on 2024-12-10; awaiting the rapporteur's report in a Chamber special committee as of the latest record (2026-09-02).
- EU-style structure: bans, high-risk list, impact assessments, generative AI duties, ANPD-led governance.
- Fines in the Senate text mirror the LGPD (R$ 50 million or 2% of revenue).
Official sources
- PL 2338/2023, Câmara dos Deputados proposition page, accessed 2026-10-19 (status read via the Chamber open data API, proposition 2487262)
- Senate-approved text as received by the Chamber (PDF), accessed 2026-10-19
- PL 2338/2023, Senado Federal (matéria 157233), accessed 2026-10-19 (Senate dates read via the Senate open data API)
This is general information, not legal advice. Last verified 19 Oct 2026 against the official source. Laws change; confirm the current position with qualified counsel before acting.