Bill No. 2338/2023 on the development, promotion and ethical and responsible use of artificial intelligence based on the centrality of the human person

Summary

Sets national rules for responsible governance of AI systems. It bans 'excessive risk' systems, such as those that exploit vulnerabilities to cause harm, assess personality traits to predict crime, public-authority universal scoring, autonomous weapons and, with exceptions, real-time remote biometric identification in public spaces. High-risk uses bring rights to explanation, contestation and human review, governance measures and algorithmic impact assessments. General-purpose and generative AI developers must assess risk and, for systemic risk, document tests. Synthetic content needs identifiers and serious incidents must be reported, under a national system (SIA) coordinated by the ANPD.

Language

English translation by Campos Thomaz Advogados (automatic machine translation by Google Translate), not by a public body Open it. The Portuguese text is authoritative.

Safety topics

Status notes

Presented in the Federal Senate by Sen. Rodrigo Pacheco on 3 May 2023, with the original ementa 'Dispõe sobre o uso da Inteligência Artificial'; the Senate calls it the 'Marco Legal da Inteligência Artificial'. The Senate's Temporary Internal Committee on Artificial Intelligence (CTIA) approved rapporteur Sen. Eduardo Gomes's substitute (Amendment No. 199) on 5 Dec 2024. On 10 Dec 2024 the Senate plenary approved the substitute, which was definitively adopted in the supplementary round without amendments, and the other Senate AI bills joined to it (PL 5051/2019, 5691/2019, 21/2020, 872/2021, 3592/2023, 210/2024 and 266/2024) were declared prejudicados. The text sent to the Chamber is the Senate's autógrafo of the substitute, signed by Senate President Rodrigo Pacheco on 31 Jan 2025; Ofício SF No. 235 of 17 Mar 2025 sent it to the Chamber of Deputies for review under art. 65 of the Federal Constitution. As adopted by the Senate, the law would enter into force 730 days after publication; the bans of art. 13, the rules for general-purpose and generative AI (Chapter IV, Section V) and the copyright rules (Chapter X, Section IV) 180 days after publication, except art. 62, which would apply at once; and Chapter IX on supervision (except the sanctions of art. 50) and Sections III and V of Chapter X on publication. Civil liability for damage caused by AI stays under the Consumer Defence Code and the Civil Code, with the burden of proof reversed when the victim is hyposufficient or proof is excessively burdensome (arts. 35 to 37), and fines are capped at R$50 million per infraction or, for private legal entities, 2% of gross revenue in Brazil (art. 50). In the Chamber the bill is subject to a plenary vote under the priority regime (art. 151, II, of the Chamber's Rules). On 4 Apr 2025 it was distributed to six committees and, because more than four had jurisdiction on the merits, a Special Committee was created; the committee was constituted on 29 Apr 2025 and Dep. Aguinaldo Ribeiro (PP-PB) was designated rapporteur on 20 May 2025. The committee approved requests for public hearings and seminars in 2025, and further bills were attached to the bill in 2025 and 2026, most recently PL 3060/2026 on 1 Sep 2026. Its Chamber status is 'Aguardando Parecer do(a) Relator(a) na Comissão Especial' (awaiting the rapporteur's report in the Special Committee); no report or substitute has been presented in the Chamber, and the last recorded action is of 2 Sep 2026. The English machine translation given as english_version renders the Senate's revised text ('Minuta') of 19 Dec 2024, the undated draft of the autógrafo with the same reference (ivb/pl23-2338rev), not the signed autógrafo of 31 Jan 2025.

Sources

Open the official source https://www.camara.leg.br/propostas-legislativas/2487262

Secondary source: https://www.camara.leg.br/proposicoesWeb/prop_mostrarintegra?codteor=2868197

Local copy

Verification

History

1. 2026-10-16verification.level: A → B — level computed from the evidence by tools/pipeline.py2026-10-16-refresh-ids-br-01-us-il-01 · Claude Code (claude-opus-5-5)
2. 2026-10-16english_version: – → {"kind": "third-party translation", "url": "https://camposthomaz.com/wp-content/ — The only English text found: a Google machine translation of the Senate's revised text of 19 Dec 2024, published by Campos Thomaz Advogados (third-party); stored in this run.2026-10-16-refresh-ids-br-01-us-il-01 · Claude Code (claude-opus-5-5)
3. 2026-10-16notes: Presented in the Senate 3 May 2023; Senate plenary approved the substitute text → Presented in the Federal Senate by Sen. Rodrigo Pacheco on 3 May 2023, with the — Rewritten from the stored pages and text: adds the CTIA approval, the Senate's final adoption and the bills declared prejudicados, the autógrafo of 31 Jan 2025, the Ofício and art. 65 review, the phased entry into force of art. 80, the liability and fine provisions, the Chamber's distribution, Special Committee and priority regime, the 2026 attachments, the absence of a Chamber report and the version the English translation renders. Removed the press reports on the timing of a vote and the remark that Chamber amendments send the bill back to the Senate, which no stored source shows, and the sentence on the excluded PBIA, which was about the research, not the bill.2026-10-16-refresh-ids-br-01-us-il-01 · Claude Code (claude-opus-5-5)
4. 2026-10-16safety_topics: ["risk-based classification", "prohibited practices", "general-purpose / frontie → ["risk-based classification", "prohibited practices", "general-purpose / frontie — Six topics, each backed by a provision: risk-based classification (arts. 12-15), prohibited practices (art. 13), general-purpose / frontier AI (arts. 29-30), human oversight (arts. 6 and 8), transparency & labelling (art. 19), incident reporting (art. 42). Dropped 'accountability & liability': the record had seven topics, one more than allowed, and the liability chapter mostly keeps the existing Consumer Defence Code and Civil Code regimes (arts. 35-36); the liability and fine provisions are now in the notes.2026-10-16-refresh-ids-br-01-us-il-01 · Claude Code (claude-opus-5-5)
5. 2026-10-16summary: Comprehensive risk-based AI framework passed by the Senate: bans 'excessive-risk → Sets national rules for responsible governance of AI systems. It bans 'excessive — Rewritten from the stored Portuguese text (arts. 1, 6, 13, 14, 18, 19, 25, 29, 30, 42 and 45), using the bill's own terms: 'excessive risk' (risco excessivo), 'high risk' (alto risco), general-purpose and generative AI (IA de propósito geral e generativa), systemic risk (risco sistêmico), synthetic content (conteúdo sintético), serious incidents (grave incidente), the SIA and the ANPD as competent authority. Moved liability and fines to the notes.2026-10-16-refresh-ids-br-01-us-il-01 · Claude Code (claude-opus-5-5)
6. 2026-10-16secondary_url: https://www25.senado.leg.br/web/atividade/materias/-/materia/157233 → https://www.camara.leg.br/proposicoesWeb/prop_mostrarintegra?codteor=2868197 — Points the record's second source at the stored bill text, so that the local copy holds the bill's own text; see local_copy_reasoning.2026-10-16-refresh-ids-br-01-us-il-01 · Claude Code (claude-opus-5-5)
7. 2026-10-16date_last_updated: 2026-09-02 → 2024-12-10 — The last new version of the text is the Senate substitute definitively adopted on 10 Dec 2024; the 2 Sep 2026 entry was an attachment of other bills, not a new version of PL 2338/2023.2026-10-16-refresh-ids-br-01-us-il-01 · Claude Code (claude-opus-5-5)
8. 2026-10-16original_title: Projeto de Lei nº 2338/2023 – Dispõe sobre o desenvolvimento, o fomento e o uso → Projeto de Lei nº 2338, de 2023 – Dispõe sobre o desenvolvimento, o fomento e o — The number in the form both chambers use in full ('Projeto de Lei nº 2338, de 2023', as on the Chamber page) followed by the ementa verbatim from the stored text; 'nº 2338/2023' does not appear in that form on any stored page.2026-10-16-refresh-ids-br-01-us-il-01 · Claude Code (claude-opus-5-5)
9. 2026-10-16title: Bill No. 2338/2023 on the development, promotion and ethical and responsible use → Bill No. 2338/2023 on the development, promotion and ethical and responsible use — Rendered from the full ementa of the stored text; the earlier title left out 'based on the centrality of the human person' (com base na centralidade da pessoa humana).2026-10-16-refresh-ids-br-01-us-il-01 · Claude Code (claude-opus-5-5)

Related records

TSE AI / deepfake election rules Brazil · In force

Bolivia AI bill (PL 178/2024-2025) Bolivia · Draft / proposed

Paraguay AI systems bill (Senate, 2025) Paraguay · Draft / proposed

Citation

National Congress of Brazil (Federal Senate, now in the Chamber of Deputies); author Sen. Rodrigo Pacheco. Bill No. 2338/2023 on the development, promotion and ethical and responsible use of artificial intelligence based on the centrality of the human person [PL 2338/2023]. 2023-05-03. https://www.camara.leg.br/propostas-legislativas/2487262. Status: Draft / proposed; verified 2026-10-16 (level B). Local copy archive/BR-01/2026-10-16/official.html retrieved 2026-10-16 18:57:08 UTC, SHA-256 1a1bcb116b49e842b0118b1b1651cf8856930c1c4f74297539a0507246bc9883. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record BR-01.

Improve this record

Know of an amendment, a status change or an error? Report news or an error on GitHub; the form opens with this record filled in.