The Electoral Attorney General's Office initiated a procedure to investigate the suspicion that big techs with generative AI systems were failing to comply with electoral rules and, within it, asked the TSE (Superior Electoral Court) to share information that it considers pertinent on the topic, based on the meetings that the court has been holding with companies.
According to the document obtained by Folha, letters were also sent by the Public Ministry to the companies responsible for ChatGPT (Open AI), Gemini (Google), Grok (X, ex-Twitter), DeepSeek, Claude (Anthropic) and Perplexity, asking them to present a preliminary statement regarding studies carried out by civil society organizations.
According to a resolution approved this year by the TSE, AI chatbots cannot rank or present suggestions for candidates and parties, nor issue positions that generate political-electoral favoritism. Reports carried out in the pre-campaign by civil society entities, however, indicate that different companies would not be meeting these requirements.
On the 27th, the deputy electoral attorney general sent a letter to the president of the TSE, Kassio Nunes Marques, requesting the sharing of information "considered pertinent" by the court sector that has been responsible for monitoring the issue, the Special Advisory for Combating Disinformation (AEED).
"It is relevant, for the investigation of the matter, that the Special Advisory for Combating Disinformation of the Superior Electoral Court (AEED/TSE) can collaborate by sharing its impressions on the understandings and work meetings that have been held" with the companies, writes the deputy attorney.
The movement occurred in a procedure initiated after notification from the Public Ministry by civil society entities in July.
Organizations linked to topics such as freedom of expression and digital rights, Conectas Human Rights, Article 19, Sleeping Giants Brasil and Justa sent a letter to the Attorney General's Office, asking for an action to be filed with the TSE against ChatGPT, Gemini, Grok and Deepseek. As a basis, a June report from the Observatório IA nas Eleições, composed of Data Privacy Brasil and Aláfia Lab, was presented.
In addition to this survey, the Public Ministry also cites in the letter a study conducted by ITS Rio (Institute of Technology and Society of Rio), "Boca de IA". From March to July, the organization carried out different rounds of analysis within this project, which has the Federal Public Ministry as a partner.
In the procedure, the Prosecutor's Office also included two journalistic reports from the end of August, one of them from the newspaper O Globo reporting that the TSE held a meeting with some of the big techs regarding compliance with the rules on AI systems. And there is also a text from the Jota website, which also carried out simulations on different AI platforms, reaching the conclusion that companies were failing to comply with the rule on ranking.
One of the challenges of monitoring this topic is the fact that each user receives different answers, based on questions and requests directed to the tools.
This is in a scenario in which this type of tool is gaining more and more relevance among voters, making the informational environment of the 2026 contest different from the previous national contest, in 2022. As Folha showed, a report prepared by 65 entities pointed to AI chatbots as one of the concerns for this year's elections. According to this analysis, these platforms did not yet have comprehensive policies to defend the integrity of elections.
The TSE required the main platforms and social networks to present detailed plans on how they prepared to comply with electoral rules. Among the bottlenecks for effective supervision, however, are the lack of court staff and also the lack of direct sanctions at the hands of the TSE for application in case of non-compliance – with the exception of companies that offer electoral advertisements (Meta and Kwai), which may be disqualified by the court from offering this service.
In view of this, the role of actors such as parties and the Public Prosecutor's Office becomes important to take legal action in the event of non-compliance.
Among the AI companies now targeted by the Public Ministry, Open AI, Google and X sent compliance plans to the TSE by the August 16 deadline. DeepSeek, Claude (Anthropic) and Perplexity do not. According to court rules, companies with more than 5 million users in Brazil were required to send.
