Policy Briefing — Brazil’s Senate Approves AI Legal Framework Bill

Executive briefing: On 24 April 2024 the Federal Senate of Brazil approved Substitute Bill 2338/2023, which establishes a comprehensive legal framework for artificial intelligence. The bill classifies AI systems by risk, mandates transparency, and introduces enforcement powers for Brazil’s data protection authority (ANPD) and other regulators. The proposal now moves to the Chamber of Deputies.

Key obligations

Risk categories. The bill defines unacceptable, high, medium, and low-risk AI, banning systems that manipulate behaviour or score citizens socially.
High-risk requirements. Providers and deployers of high-risk AI must implement risk management programmes, maintain technical documentation, conduct impact assessments, and ensure human oversight.
Transparency duties. Users must be informed when interacting with AI, and providers must disclose training data characteristics, limitations, and safeguards.
Fundamental rights protections. The bill guarantees rights to explanation, contestation, and human review for individuals affected by AI decisions.
Sanctions. Violations can trigger fines up to 50 million Brazilian reais or 2% of a company’s Brazilian revenues, alongside daily penalties and product suspension.

Governance structure

Central authority. The bill proposes a national AI authority, with the ANPD initially assuming oversight until a dedicated regulator is established.
Sector regulators. Existing regulators (e.g., Central Bank, Anatel, SUSEP) will enforce sector-specific provisions in coordination with the AI authority.
Regulatory sandbox. The framework enables sandboxes to test innovative AI solutions under regulator supervision.

Program actions

Legislative monitoring. Track deliberations in the Chamber of Deputies and potential amendments affecting scope or enforcement timelines.
Risk inventory. Classify AI systems in Brazil against the proposed risk tiers and document controls, audits, and human oversight.
Transparency readiness. Prepare user disclosures, contestation workflows, and evidence for ANPD inspections.
Cross-framework alignment. Map Brazil’s proposed obligations to EU AI Act and ISO/IEC 42001 requirements to streamline compliance.

Sources

Zeph Tech guides organisations through Brazil’s emerging AI legal framework by pairing risk classification, documentation, and transparency programmes with global governance standards.