Veronyka Gimenes wants you to know AI can be used to fight back. In 2025, Gimenes and Amanda Claro, co-founders of the Brazilian advocacy group Código Não Binário, created TybyrIA, an open-source AI model for detecting anti-LGBTQIA+ hate speech. The model allowed them to categorize evidence at scale for a lawsuit against social media companies.The motivation for TybyrIA traces back to 2024, when a clip of Entre Amigues—a queer podcast produced by Código Não Binário—went viral. The clip, which discussed the Brazilian transmasculine identity “boyceta,” triggered an anti-trans hate campaign. Far-right figures mockingly shared the clip with their audiences. Gimenes says the resulting round-the-clock bombardment of comments and messages took a severe toll on the team’s mental health.Gimenes realized the scale of hate could provide the data for a “counter-attack.” But after manually sorting malicious comments proved too psychologically taxing, she turned to AI. Fine-tuning an open-source model allowed her to catalogue hate speech at a previously unimaginable scale—evidence Gimenes later used to file a public civil action lawsuit against Meta, Google, X, and ByteDance. The suit seeks roughly $20 million in damages and demands platforms take more proactive measures to curb anti-LGBTQIA+ hate on the platforms. According to court filings viewed by TIME, the companies argue that their policies already prohibit discrimination based on sexual orientation and gender identity, and that there are mechanisms to remove content that violates those policies in place. (Meta declined to comment. Google, X and ByteDance did not respond to requests for comment.)Gimenes now works to help others understand how they can use AI critically, rather than altogether rejecting the technology. Her online course with Código Não Binário, which has already seen more than 1,700 enrollments, teaches people how they can leverage and adapt open-source technologies.