A coordinated far-right hate campaign, amplified by social media algorithms, led the Brazilian NGO Código Não Binário to develop an artificial intelligence tool to detect anti-LGBT+ hate speech, and to take major tech companies to court over digital violence.
The organization filed a public civil action, a legal instrument under Brazilian law used to defend collective rights, against Meta (owner of Facebook, Instagram and WhatsApp), Google (YouTube), X and ByteDance. The lawsuit argues that these platforms jointly enable and profit from the circulation of anti-LGBT+ hate. The case was filed with support from civil society organizations including the Brazilian Institute of Transmasculinities (Ibrat), the National Forum of Black Trans Women and Men (Fonatrans), and feminist magazine AzMina.
The initiative emerged after a clip from Entre Amigues, a podcast produced by Código Não Binário, went viral in May 2024. The episode discussed the term “boyceta”, used in the outskirts of São Paulo to refer to transmasculine people. The clip became a trending topic on X (Elon Musk‘s social media platform) for two days, largely driven by hate speech, and was pushed by TikTok’s algorithm to millions of users.
To analyze the attack without exposing team members to prolonged psychological harm, the NGO developed a data-extraction methodology. Two thousand comments were manually classified, and an AI system was then trained to analyze the remaining content. The resulting technical report became the basis of the lawsuit.
The tool was named TybyrIA, in reference to Tybyra of Maranhão, considered the first documented victim of anti-LGBT+ violence during Brazil’s colonial period, executed by Portuguese authorities. The AI is free, open-source and publicly available on Código Não Binário’s website.
“If data is the ‘new oil,’ we took that seriously,” said Veronyka Gimenes, one of the NGO’s directors, in an interview with Radio BdF. “We transformed hate into value, into power, to fight the circulation of hate through AI.”
Gimenes stressed the importance of demystifying artificial intelligence. Unlike large commercial language models, TybyrIA is a targeted analytical tool. “This isn’t ChatGPT,” she said. “It has nothing to do with massive, profit-driven systems that require city-sized data centers. Confusing these things pushes people away from AI and reflects a lack of understanding of the field.”
The AI’s findings showed that hate comments receive significantly more engagement than average posts. According to the analysis, platform algorithms tend to amplify intense emotions, including anger, effectively rewarding hateful content with greater visibility.
“Our legal system isn’t yet equipped to regulate social networks at the speed these changes happen,” said Amanda Claro, another director of Código Não Binário. “This initiative works to fill that gap.”
Based on the data and analysis, the NGO launched what it describes as an unprecedented legal action in Brazil, targeting all major platforms simultaneously. The argument is that they operate as an interconnected system rather than isolated services.
“They form and function as a deregulated system that allows hate to circulate freely and rewards it,” Claro said. “These platforms aren’t neutral spaces. Hate operates through their engagement rules.”
