Skip to content
AI Policy Generator

ESWhat applies

Brazil · Bill

Brazil's AI framework bill

BillPL 2338/2023

Summary

It would create a risk-based AI framework inspired by the EU regulation. It bans excessive-risk uses; for high-risk uses (employment, credit, insurance, education, health, biometrics) it requires algorithmic impact assessments, human oversight, documentation with bias testing and public registration. It also requires labeling synthetic content and paying copyright holders for training.

Who it applies to

If passed, to developers, distributors and deployers of AI systems in Brazil.

Penalties

Bill: fines of up to 2% of revenue in Brazil, capped at R$50 million per violation, doubled for repeat offenses; suspension in serious cases.

Official name
Projeto de Lei nº 2.338, de 2023 – Dispõe sobre o desenvolvimento, o fomento e o uso ético e responsável da inteligência artificial
Official reference
Aprobado por el Senado el 10-dic-2024; en la Cámara de Diputados desde el 17-mar-2025 (Comisión Especial; "Aguardando Parecer" al 27-oct-2026)
Regulator
ANPD as coordinator of the proposed National AI Regulation and Governance System
Official source
Check the sourceHow we verify sources

Key dates

Obligations and dates

Obligations table
ObligationArticle or citationStatusDateWho it applies toVerification
Review of excessive-risk usesPL 2338/2023 (texto del Senado), art. 13BillAlready applicableRecommended preparation for any organization using AI in Brazil.Pending confirmation in the official source (last checked: November 4, 2026)
Preliminary risk assessment (voluntary)PL 2338/2023 (texto del Senado), art. 12BillAlready applicableRecommended preparation for any organization using AI.Pending confirmation in the official source (last checked: November 4, 2026)
Algorithmic impact assessment and human oversight for high riskPL 2338/2023 (texto del Senado), arts. 8, 14, 18, 25 y 44BillAlready applicableIf you use AI for employment, credit, insurance, education, health, essential services or biometrics.Pending confirmation in the official source (last checked: November 4, 2026)
Synthetic content labeling and copyright in trainingPL 2338/2023 (texto del Senado), arts. 19 y 62 a 65BillAlready applicableIf you generate synthetic content or train models on copyrighted works.Pending confirmation in the official source (last checked: November 4, 2026)

Review of excessive-risk uses

PL 2338/2023 (texto del Senado), art. 13Bill

Already applicable

When it relates
Recommended preparation for any organization using AI in Brazil.
What the rule provides
  • Confirm no system induces harmful behavior, exploits vulnerabilities, assesses personality traits to predict crime, facilitates child sexual abuse material or performs real-time remote biometric identification outside the bill's exceptions.
  • Social scoring would be banned only for public authorities; emotion recognition is treated as a high-risk use (art. 14, XI), not a banned one.
Typical evidence
  • Signed checklist.
Related base documents
Verification
Pending confirmation in the official source (last checked: November 4, 2026)Per the status published by the Chamber of Deputies (open data) as of Oct 27, 2026: "Awaiting report", priority procedure; the text may change. Content checked against the text passed by the Senate and sent to the Chamber, art. 13 (banned uses) and art. 14, XI. The announcement of a vote after the elections comes from secondary sources.

Preliminary risk assessment (voluntary)

PL 2338/2023 (texto del Senado), art. 12Bill

Already applicable

When it relates
Recommended preparation for any organization using AI.
What the rule provides
  • Carry out a preliminary assessment of each system's risk level before using or selling it; the Senate text makes it voluntary and treats it as good practice.
Typical evidence
  • System register with classification.
Verification
Pending confirmation in the official source (last checked: November 4, 2026)Per the status published by the Chamber of Deputies (open data) as of Oct 27, 2026: "Awaiting report", priority procedure; the text may change. Content checked against the text passed by the Senate and sent to the Chamber, art. 12 ("may carry out a preliminary assessment", treated as good practice). The announcement of a vote after the elections comes from secondary sources.

Algorithmic impact assessment and human oversight for high risk

PL 2338/2023 (texto del Senado), arts. 8, 14, 18, 25 y 44Bill

Already applicable

When it relates
If you use AI for employment, credit, insurance, education, health, essential services or biometrics.
What the rule provides
  • Prepare the algorithmic impact assessment (art. 25), human oversight that allows people to understand, interpret, decide and intervene (art. 8) and technical documentation with bias testing (art. 18).
  • The authority would keep a public database of high-risk systems' impact assessments (art. 44); it helps to have a public version ready.
Typical evidence
  • File per high-risk system.
Verification
Pending confirmation in the official source (last checked: November 4, 2026)Per the status published by the Chamber of Deputies (open data) as of Oct 27, 2026: "Awaiting report", priority procedure; the text may change. Content checked against the text passed by the Senate and sent to the Chamber, arts. 8, 14, 18, 25 and 44. The announcement of a vote after the elections comes from secondary sources.

Synthetic content labeling and copyright in training

PL 2338/2023 (texto del Senado), arts. 19 y 62 a 65Bill

Already applicable

When it relates
If you generate synthetic content or train models on copyrighted works.
What the rule provides
  • Label AI-generated content.
  • Respect rights holders' opt-out and remuneration for use of works in training.
Typical evidence
  • Labeling policy and training source log.
Verification
Pending confirmation in the official source (last checked: November 4, 2026)Per the status published by the Chamber of Deputies (open data) as of Oct 27, 2026: "Awaiting report", priority procedure; the text may change. Content checked against the text passed by the Senate and sent to the Chamber, art. 19 (identifier in synthetic content) and arts. 62-65 (disclosure, opt-out and remuneration for rights holders). The announcement of a vote after the elections comes from secondary sources.

Related rules

Regulatory change alerts by country

Pick the countries you care about and we will email you when we change a rule or an application date for those countries in our registry.

  • Changes to the rules of the countries you pick, when we add them to the registry.
  • New application dates.
  • No more than one email a week. Unsubscribe at any time.

It is a free, informational notice, not a legal monitoring service: it does not necessarily catch every change or within a set time.

Countries · Pick one or more.
US states
Email language
Manage my alerts

Find out what applies to you (free)