The measure carries a daily fine of BRL 50,000 for non-compliance, and the ANPD cites about 102 million active Facebook users in Brazil.
As of November 17, 2026, the ANPD note records the measure and the daily fine. No ANPD source read records a lifting, a modification or a fine actually imposed, so none is stated. The page belongs to the independent reference on AI training data licensing, and DataPacts is not affiliated with the ANPD or Meta.
Dates, decision numbers and quotations come from the ANPD news note of 2024-07-02, read on November 17, 2026, and how figures are sourced and dated is set out on the methodology page.
The ANPD Preventive Measure Against Meta at a Glance: Authority, Documents, Fine and Status
The ANPD ordered the immediate suspension in Brazil of Meta's updated privacy policy for AI training on July 2, 2024, and attached a daily fine of BRL 50,000 for non-compliance. The table lists the facts of the measure as of 2026-11-17.
| Field | Value | Source |
|---|---|---|
| Measure | Medida Preventiva against Meta (Facebook, Messenger, Instagram), ANPD | ANPD note |
| Authority | Autoridade Nacional de Proteção de Dados (ANPD), Conselho Diretor | ANPD note |
| Date | 2024-07-02 | ANPD note |
| Decision documents | Voto n. 11/2024/DIR-MW/CD; Despacho Decisório n. 20/2024/PR/ANPD | ANPD note |
| Subject | updated privacy policy (in force 2024-06-26, as the ANPD describes it) allowing public posts and shared content to train generative AI | ANPD note |
| Order | immediate suspension in Brazil of the policy and of the processing | ANPD note |
| Fine | daily fine of BRL 50,000 for non-compliance | ANPD note |
| Scale cited | about 102 million active Facebook users in Brazil (ANPD's figure) | ANPD note |
| Status | measure recorded; later course not verified | ANPD note; checked 2026-11-17 |
The Four Preliminary Findings Behind the Suspension
The ANPD relied on four preliminary findings: an inadequate legal basis, missing clear information, excessive limits on data subjects' rights, and processing of children's and adolescents' data without safeguards.
The authority opened an inspection process of its own motion (de ofício), and its preliminary analysis found a risk of serious harm that is hard to repair. The Conselho Diretor approved the Voto in a deliberative circuit.
The policy applies to the "Meta Products" (Facebook, Messenger and Instagram) and covers publicly available information and content shared by users.
The ANPD states that the policy update entered into force on June 26, 2024, which is the date used in the table above, and it places the measure within an inspection process that is to assess the conduct in detail.
The table gives each finding as the ANPD words it, as of 2026-11-17.
| Preliminary finding | What the ANPD says | Source |
|---|---|---|
| Legal basis | legitimate interest judged inadequate in preliminary analysis; not usable for sensitive data; legitimate expectations, purpose and necessity must be weighed | ANPD note |
| Transparency | no clear, precise and easily accessible information on the policy change and the processing | ANPD note |
| Data subject rights | excessive and unjustified obstacles to information and to objecting | ANPD note |
| Children and adolescents | photos, videos and posts could be used without the safeguards and risk mitigation that the LGPD requires in the child's best interest | ANPD note |
On the third finding, the ANPD says that users could object but met "excessive and unjustified obstacles" to information and to exercising the right.
All four are preliminary findings, not a final decision, and the note does not describe any of them as an established violation. The findings do not include any statement about Meta's reply or about the content of its training data, because the note carries none.
Why the ANPD Judged Legitimate Interest Inadequate
The ANPD considered Meta's legal basis, its legitimate interest, inadequate in the preliminary analysis because that basis cannot be used for sensitive personal data and requires attention to data subjects' legitimate expectations and to the principles of purpose and necessity.
The ANPD does not rule on which legal basis would be acceptable, and it does not call legitimate interest unlawful for AI training in general. Its reasoning is limited to the policy before it and to the facts in the preliminary analysis.
The ANPD reasons that content on these platforms is generally shared for relationships with friends, a close community or businesses of interest. Users would therefore not necessarily expect all of it, including content shared many years earlier, to train AI systems that were not yet implemented when it was shared.
In the ANPD's reading, the purpose and necessity principles ask what the processing is for and whether it needs the data in that form, and the legitimate expectations of the person who posted weigh in that test.
The same legal basis is contested in a different forum in the EU, and the chronology of Meta AI training on EU user data is kept on its own page.
What a Preventive Measure Is and What It Leaves Undecided
A Preventive Measure is an instrument within the competence of the ANPD's directors, used to protect data subjects' rights and prevent serious or hard-to-repair harm, and it assesses a company's conduct only preliminarily.
It may issue in urgent cases without a prior hearing of the interested party, and it may carry a daily fine for non-compliance. The detailed assessment follows later in the inspection process run by the ANPD's technical area.
The note therefore decides a suspension and a fine for non-compliance, and it leaves the final assessment to that inspection process. No final decision is recorded in the note, and the findings are preliminary as the note records them.
A suspension of this kind stops the processing for the time being; it does not by itself decide that any past processing was unlawful. The sanction scale of the LGPD and the other instruments of Brazilian law are set out in the page on rules for licensing data for AI training in Brazil.
What the Measure Signals to Platforms and Data Holders
The measure shows that a regulator can stop AI training on user content before any final finding, and it sets neither a price nor a general rule for licensing data. The note's three themes are practical points for any platform that changes its terms:
- transparency about a policy change and the processing it allows;
- the practical ease of objecting to that processing;
- safeguards for the data of children and adolescents.
The measure concerns a platform's own users and is not a statement about licensed data sold to third parties. It also does not address copyright, which the Brazil page treats separately from the LGPD. What platforms' terms say is covered on the page about services that train on their users.
A holder of user-generated or customer data meets the consent and confidentiality questions on the page about customer consent before licensing data. The ANPD note makes no prediction about further action, and neither does this page.
Where the ANPD Measure Sits Among Other Regulator and Court Records
The ANPD measure is one of the regulator actions in the case tracker that concern personal data used to train AI, alongside the Italian Garante's 2024 decision against OpenAI and the EU proceedings over Meta.
The case tracker of AI training data cases lists every tracked case with docket and status. The table names four related matters as of 2026-11-17.
| Matter | Forum and reference | Status | Source |
|---|---|---|---|
| Meta AI training on EU user data | DPC, noyb, OLG Köln 15 UKl 2/25 | interim application rejected 2025-05-23; main proceedings not found (as of 2026-11-09) | case page |
| Italian Garante v. OpenAI | Garante provvedimento n. 755 of 2024-11-02; Tribunale di Roma n. 4153/2026 | OpenAI's opposition upheld, judgment published 2026-03-18 (as of 2026-11-10) | case page |
| Kadrey v. Meta | N.D. Cal. 3:23-cv-03417-VC | training claim decided for Meta on the record 2025-06-25 (copyright, a different claim) | case page |
| EDPS v SRB | CJEU C-413/23 P (pseudonymised data) | decided 2025-09-04; T-557/20 referred back to the General Court | case page |
The rows differ in forum, legal question and status, so no row is a precedent for another. Meta's own licensing and acquisition programs sit on the page about Meta's data programs and are not part of this measure.
More About the ANPD Measure
This page is general information about a regulator measure, not legal advice for any specific matter.
Verified Versus Not Verified in the ANPD Record
The ANPD note records the suspension, the four findings and the daily fine; it does not record a lifting, a compliance plan, a modification or an imposed fine, and no other primary source read does either.
| Recorded in the ANPD note | Not verified |
|---|---|
| suspension of the policy in Brazil, 2024-07-02 | whether and when the measure was lifted or modified |
| the four preliminary findings | any fine actually imposed |
| the daily fine of BRL 50,000 for non-compliance | the outcome of the inspection process |
The page of the note shows publication and last modification on 2024-07-02, with no later update, as read on 2026-11-17.
Frequently Asked Questions About Brazil and Meta AI Training
What did the ANPD order against Meta?
The ANPD ordered the immediate suspension in Brazil of Meta's updated privacy policy for AI training, with a daily fine of BRL 50,000 for non-compliance. The order is the preventive measure of 2024-07-02.
Was the ANPD order lifted?
A lifting is not verified here. No ANPD source read records one, and the note was not updated after 2024-07-02 (checked 2026-11-17).
Did the ANPD fine Meta?
The note sets a daily fine for non-compliance, and no imposed fine is recorded. The BRL 50,000 figure is the rate of the daily fine, not an amount paid, and the note gives no number of days.
Does the measure apply outside Brazil?
The measure suspends the policy in Brazil only. The EU record is a separate matter with its own forum and status.