ANPD, MPF, and Senacon order immediate measures against X to remedy failures in an AI tool and prevent the generation of improper content

ANPD, MPF, and Senacon order immediate measures against X to remedy failures in an AI tool and prevent the generation of improper content

On February 11, 2026, the Brazilian National Data Protection Authority (“ANPD”), Brazilian Federal Public Prosecutor’s Office (“MPF”), and National Consumer Secretariat (“Senacon”) ordered the X platform to immediately implement technical and operational measures to remedy failures identified in its artificial intelligence tool. The decision, formalized through Decision Order SEI/ANPD No. 0247250, was adopted after the authorities concluded that the measures initially presented by the company failed to demonstrate sufficient effectiveness in preventing the generation and dissemination of sexualized content involving real individuals—including children and adolescents—without consent.

The order represents the most recent development in proceedings initiated on January 14, 2026, when ANPD’s General Coordination for Enforcement received Representation No. 01-2026 (0238713), submitted by Federal Deputy Erika Hilton. The submission identified potential irregularities in the operation of the artificial intelligence system Grok, particularly concerning the creation of synthetic images depicting identified or identifiable individuals in sexualized contexts.

As part of the preliminary assessment, ANPD issued, on January 20, Technical Note No. 0239948, which examined the legal risks associated with the generation of deepfakes involving real people. The document advances the interpretation that the creation of synthetic content based on elements that enable the identification of an individual constitutes the processing of personal data under Brazil’s General Data Protection Law (Lei Geral de Proteção de Dados Pessoais – LGPD). The Technical Note further emphasizes that, in such cases, duties of security, prevention, and accountability apply, with heightened rigor where data relating to children and adolescents are concerned.

Also on January 20, ANPD, the Brazilian Federal Public Prosecutor’s Office, and the National Consumer Secretariat issued a Joint Recommendation to the platform, underscoring the need to adopt structural safeguards capable of preventing the use of the tool for the production of sexualized content involving real individuals. The document stresses that artificial intelligence systems must incorporate preventive mechanisms from the design stage onward and is grounded not only in the LGPD, but also in the Statute of the Child and Adolescent (Estatuto da Criança e do Adolescente – ECA), the Consumer Protection Code (Código de Defesa do Consumidor – CDC), and the Brazilian Internet Civil Rights Framework (Marco Civil da Internet – MCI). This reflects an integrated approach to personal data protection, the safeguarding of human dignity, and consumer protection in the digital environment.

Following the review of the information submitted by the company, the authorities concluded that there was no consistent technical evidence demonstrating that the identified failures had been effectively remedied. Accordingly, the Decision Order required the immediate adoption of mechanisms capable of blocking the generation of new content of this nature, covering all versions and functionalities of the tool, as well as the submission of detailed documentation demonstrating the effectiveness of the measures implemented.

The case highlights the strengthening of institutional oversight over high-risk generative technologies and reaffirms the requirement that developers and operators of artificial intelligence systems adopt concrete, verifiable, and proportionate measures to prevent violations of fundamental rights, particularly in cases involving the comprehensive protection of children and adolescents.

Technical Note No. 0239948, the Joint Recommendation, and the response issued by ANPD may be accessed through the following links: SEI/ANPD – 0239948 – Technical Note; Joint Recommendation; Response to the Joint Recommendation.

The procedural history can be accessed at: Case No. 00261.000178/2026-27.

Note: For quick release, this English version is provided by automated translation without human review.