By Sagar Shankaran, Founder of CallSphere
ANPD's 2024 Generative AI study and 2026 sandbox guidance turn LGPD principles into concrete voice and chat compliance steps. PL 2338 is moving through Congress. Penalties reach R$50M per infraction.
Key takeaways
Brazil's LGPD has been in force since 2020 and ANPD has matured into a peer of CNIL and ICO. The 2024 Generative AI study, the AI sandbox in 2026, and the pending PL 2338 give voice and chat a clear playbook.
Lei Geral de Proteção de Dados (LGPD, Lei 13.709/2018) governs personal-data processing in Brazil. Ten lawful bases echo the GDPR plus a "legitimate interests" basis with required balancing. Sensitive personal data — racial origin, religion, political opinion, union membership, health, sex life, genetic, biometric — needs specific consent or a narrow exception. ANPD's enforcement powers include warnings, daily fines, suspension, and fines up to 2% of group turnover in Brazil capped at R$50M per infraction. Article 20 grants a right to review of decisions made solely on automated processing that affects interests, including profiling — analogous to GDPR Article 22 but explicitly retained even after the Constitutional Court limited the original full-explainability text.
ANPD's November 2024 study Inteligência Artificial Generativa (Technological Radar) sets expectations: legal basis must be documented per processing operation; legitimate interests requires a three-step test; data subject rights apply to outputs; cross-border transfers need a basis under Articles 33–36. Brazil's Plano Brasileiro de Inteligência Artificial 2024–2028 funds Portuguese-language foundation models and a regulatory sandbox. PL 2338/2023, the AI Bill, is moving through the Senate and would establish a risk-based AI regime.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent in your browser — 60 seconds, no signup.
Voice agents serving Brazilian residents need a documented LGPD lawful basis per processing operation, a privacy notice in Portuguese with the controller's identity and DPO contact, and an Article 20 review pathway for any solely automated decision that affects the data subject. Sensitive-data processing — biometric voiceprints, health intent — needs specific consent. Cross-border transfers go via Standard Contractual Clauses (Brazilian model adopted in 2023), specific consent, or adequate jurisdiction. ANPD-reportable breaches must move quickly. Generative-AI deployments document training-data provenance per the 2024 ANPD study.
CallSphere — 37 agents, 90+ tools, 115+ DB tables, 6 verticals, 50+ businesses, 4.8/5, HIPAA and SOC 2 aligned — ships Portuguese-language voice notices, an LGPD lawful-basis matrix per workflow, an Article 20 review trigger that hands solely-automated decisions to a human, and ANPD-aligned breach-response runbooks. SCC-equivalent transfer documents are generated automatically. The 2024 ANPD GenAI checklist is built into the model-onboarding flow. Pricing $149 / $499 / $1,499; 14-day trial; 22% affiliate; see /pricing and /contact.
flowchart LR
A[BR Caller] --> B[Voice Agent\nPT-BR]
B --> C[LGPD Basis Map]
B --> D[Art 20 Review]
D --> E[Human Reviewer]
B --> F[ANPD Breach\nRunbook]
F --> G[Cross-Border\nSCC]
Is consent the default basis? No — LGPD has ten bases. Consent is one. Contractual necessity, legitimate interests, and legal obligation are common alternatives.
Are voice biometrics sensitive? Yes — biometric data for unique identification is sensitive and requires specific consent or another Article 11 exception.
Still reading? Stop comparing — try CallSphere live.
CallSphere ships complete AI voice agents per industry — 14 tools for healthcare, 10 agents for real estate, 4 specialists for salons. See how it actually handles a call before you book a demo.
Does ANPD recognise GDPR DPIAs? A GDPR DPIA is a strong starting point but should be re-papered as an LGPD RIPD.
Can we skip Portuguese in B2B? The notice should be in the language the data subject understands; Portuguese is the safe default.
Is there a cure period? ANPD often issues warnings before fines but is not obligated to. Treat warnings as fast-onset.
Written by
Sagar Shankaran· Founder, CallSphere
Sagar Shankaran is the founder of CallSphere, where he builds production AI voice and chat agents deployed across healthcare, hospitality, real estate, and home services. He writes about agentic AI, LLM engineering, and shipping voice agents that handle real calls in production.
See how AI voice agents work for your industry. Live demo available -- no signup required.
How generative AI produces verified dbt models for data migration — from scratch and incrementally — with SME validation and strict data governance.
A founder's guide to the female voice generator landscape: AI female voices, Japanese voices, robot voices, and how CallSphere ships 57+ voices live.
MOS 4.3+ is the band where AI voice feels human. Drop below 3.6 and conversations break. Here is how to measure, improve, and alert on MOS in production AI voice using G.711, Opus, and the underlying packet loss / jitter / latency math.
Why Claude real estate is reshaping voice and chat automation, with concrete patterns for real estate AI in production deployments. A field-tested view from production teams shi...
How leaders should think about Claude legal contracts — adoption patterns, ROI, competitive dynamics, and what legal tech AI means for the next 12 months.
How leaders should think about MCP registry economy — adoption patterns, ROI, competitive dynamics, and what AI marketplace means for the next 12 months.
© 2026 CallSphere LLC. All rights reserved.