Regulation of New Technologies
AI governance for companies operating in Brazil

In this article
Brazil does not yet have a dedicated AI statute: the main bill, PL 2338/2023, was approved by the Federal Senate in December 2024 and is still pending in the Chamber of Deputies as of October 2026. Companies using AI in Brazil are nonetheless already subject to binding rules, chiefly the data protection law (LGPD), the Consumer Protection Code, copyright and anti-discrimination rules. A sound AI governance program built on these laws, and designed to absorb the future statute, is the prudent path.
Key points
- PL 2338/2023 follows a risk-based approach and will return to the Senate after the Chamber votes it.
- The LGPD already grants a right to request review of decisions based solely on automated processing.
- Consumer, copyright, labor and anti-discrimination rules apply to AI use today.
- Governance starts with an inventory of AI systems and a risk classification.
- Contracts with AI vendors are a key control point for data, IP and liability.
Is there an AI law in Brazil?
Not yet. PL 2338/2023 was approved by the Senate in December 2024 and is under review by a special committee in the Chamber of Deputies. If the Chamber amends it, the bill goes back to the Senate before presidential sanction. The Senate text classifies AI systems by risk, prohibits certain uses considered of excessive risk, imposes obligations on high-risk systems and provides for a coordinated supervisory system. Because the text may still change, companies should track it but not treat its details as law.
Which existing laws already apply to AI?
- LGPD (Law 13,709/2018): any AI that processes personal data needs a legal basis, transparency and security. Article 20 gives individuals the right to request review of decisions taken solely on the basis of automated processing that affect their interests, including profiling.
- Consumer Protection Code (Law 8,078/1990): chatbots, recommendation engines and automated credit or pricing decisions must respect information duties and cannot be abusive. Suppliers are, as a rule, strictly liable for defective services.
- Copyright Law (Law 9,610/1998): training data and generated outputs raise authorship and licensing questions that remain unsettled.
- Labor law and anti-discrimination rules: AI used in hiring, evaluation or monitoring must avoid discriminatory effects.
- Sector regulation: financial, health and telecom regulators supervise technology risk in their sectors.
How to build an AI governance program
- Inventory: list every AI system in use or development, including tools embedded in third-party software.
- Classify risk: consider the impact on individuals (credit, employment, health, access to services), data sensitivity and the degree of automation.
- Assign ownership: define who approves new uses, usually a committee involving legal, technology, data protection and the business.
- Assess impact: for high-impact uses, document a data protection impact assessment and a bias and accuracy review.
- Ensure human oversight: define when a person must review or can override the system.
- Be transparent: tell users when they interact with AI and explain automated decisions in clear language.
- Monitor and log: track performance, incidents and complaints, and keep records that show diligence.
What should AI vendor contracts include?
Most companies use AI through vendors. The contract is where risk is allocated. Key provisions include: whether your data may be used to train the vendor’s models; data location and international transfers (subject to the ANPD’s standard contractual clauses under Resolution CD/ANPD 19/2024); confidentiality; ownership and licensing of outputs; IP infringement warranties and indemnities; security and incident notification; audit rights; and exit and data return.
Special points for foreign groups
Global AI policies, often drafted with the EU AI Act in mind, are a useful starting point, but they need local adaptation. Brazilian law has its own rules on automated decisions, consumer liability and data transfers. Privacy notices and user-facing explanations must be in Portuguese, and the Brazilian subsidiary should have clear authority to suspend a tool that creates local legal risk.
Frequently asked questions
Does the ANPD supervise AI?
The ANPD supervises the processing of personal data, which covers a large share of AI uses. Its role under the future AI statute depends on the final text of PL 2338/2023.
Can employees use public generative AI tools?
Only under a clear policy. Entering confidential or personal data into public tools may breach confidentiality duties and the LGPD.
Should we wait for the AI law before acting?
No. Current laws already apply, and a governance program built now can be adjusted when the statute is enacted.
How Chambarelli Advogados can help
Our Regulation of New Technologies team and our Foreign Desk design AI governance programs, review vendor contracts and adapt global policies to Brazilian law. Book a meeting with one of our partners.
Informational content, updated October 2026. It does not replace advice on your specific case.
