Is this checklist enough for LGPD compliance?

It covers the AI-traffic obligations an ANPD review reaches for: accountability, legal basis, international transfer, automated decisions, security, and breach reconstruction. It does not replace a full LGPD program covering collection notices, retention, the encarregado role, and data subject rights across all channels. Treat it as the AI-specific layer on top of an existing program.

What matters most under the LGPD for AI right now?

Article 6 accountability, Article 20 automated decisions, and the Articles 33 to 36 transfer rules, because those are where AI traffic differs most from ordinary processing and where the ANPD's 2025 technical note signals attention. The transfer rules bite immediately, since most model endpoints sit outside Brazil.

Does PL 2338 change this checklist?

If enacted, PL 2338/2023 would add a risk-tiered AI framework with its own obligations on top of the LGPD, so the checklist would grow rather than shrink. The identity, classification, transfer, and per-decision records built now are the same artifacts a risk-tiered regime would demand, so the work is not wasted while the bill moves through the Chamber of Deputies.

Brazil LGPD AI Compliance Checklist: Ten Items Before a Prompt Carries Personal Data

The LGPD (Law 13.709/2018) already governs every AI system that processes personal data in Brazil, and the Autoridade Nacional de Proteção de Dados (ANPD) enforces it now, ahead of the dedicated AI bill PL 2338/2023 that the Senate approved on 10 December 2024. This checklist follows the order an ANPD review tends to take: identity and accountability records first, then legal basis, then international transfers, then automated decisions and breach readiness.

Each item pairs the action with the evidence it produces, because the LGPD's accountability principle (prestação de contas) demands demonstration, and a checklist of intentions demonstrates nothing.

1. Bind identity to every AI request

Route AI traffic so a named user or agent, not a shared service key, is attached to each prompt. Evidence comes from a per-request record showing which identity sent which prompt to which model, which underpins the Article 6 accountability principle.

2. Classify the personal data in the prompt

Evaluate the data classification of the context window before it leaves, separating ordinary personal data from Article 11 sensitive data. Evidence is a per-request classification decision, supporting the legal-basis obligations of Articles 7 and 11.

3. Enforce legal basis by role

Block prompts that carry sensitive personal data when the caller's role lacks a lawful basis for that processing. Evidence is the per-request policy decision showing what each role was permitted to send.

4. Record the destination country of every transfer

Log the model endpoint and its country for each request carrying personal data. Evidence is a destination log that answers the Articles 33 to 36 international-transfer question directly.

5. Enforce transfer rules at the boundary

Route prompts carrying regulated classifications only to endpoints in permitted countries or under an approved mechanism per ANPD Resolution CD/ANPD 19/2024. Evidence is the enforcement log, and AI data residency controls are how this becomes real rather than contractual.

6. Produce a per-decision record for automated decisions

For any AI output that affects a data subject's interests, write a record binding the input prompt to the output and the applied logic. Evidence is decision lineage, the operational backing for the Article 20 right to review, mirroring GDPR Article 22 handling.

7. Support the Article 20 review right

Provide clear information on the criteria and procedures of automated decisions, and flag decisions that must support review. Evidence is the review-flag record plus the criteria documentation tied to actual decisions.

8. Replace static API keys with per-role, per-route policy

Remove shared credentials that grant permanent full model access, satisfying the Article 46 security obligation. Evidence is an access policy plus the enforcement log showing which roles reached which models.

9. Keep the record independent of the application

Ensure the compliance log is written by a system separate from the one making the AI call, so it survives selective logging, suppression, and crash-loss. Evidence is a signed, tamper-evident log outside application custody, as argued in You Own the AI Liability, Not the Vendor.

10. Pre-build the Article 48 breach path

Confirm you can reconstruct an AI-channel exposure end to end, because Article 48 requires notifying the ANPD and data subjects of incidents that may cause risk. Evidence is a query against the independent request log that produces a timeline. My honest take: the notification clock does not pause while you go looking for the log.

How the checklist maps to the controls

Items 1 through 10 resolve to a handful of enforcement points at the request boundary, laid out in the LGPD controls mapping. The audit-evidence view covers what each artifact looks like when the ANPD asks. Together they turn this list from tasks into a running system with evidence attached.

DeepInspect

Nine of these ten items produce evidence at the same place: the boundary between your users or agents and the LLM APIs they call. DeepInspect sits inline at that boundary. It binds identity to each request, classifies the prompt, enforces legal-basis and destination policy, flags automated decisions for review, and writes a signed per-decision record before the response returns. The checklist becomes a query rather than a scramble.

For a controller facing closer ANPD scrutiny of Article 20 and the transfer rules, the accountability principle in item 1 and the review right in item 7 rest on records that actually exist. Book a technical deep dive at deepinspect.ai.