AI risks for LGBTQ+ communities ^{CEFR B1}

AI is becoming more common and private investment has risen over the past decade. A global Ipsos survey found 55 percent of respondents thought AI-powered solutions had more benefits than drawbacks. Companies often highlight AI's efficiency and ease of use, but concern about risks remains.

Concerns are strong in LGBTQ+ communities. Wired reported that some image-generation tools, such as Midjourney, produced reductive and harmful images of LGBTQ+ people. Internet data can carry stereotypes and AI models trained on that data may repeat them. UNESCO examined assumptions behind several large language models and found tools like Meta's Llama 2 and OpenAI's GPT-2 created negative content about gay people more than half of the time in simulations.

Forbidden Colours said automatic gender recognition (AGR) uses facial features or voice to guess gender, which cannot show how people understand their own gender. Politico Europe reported that Viktor Orbán allowed AI-enabled biometric monitoring at local Pride events, a move that lets authorities surveil artists, activists and ordinary citizens. European Union institutions are reviewing that policy. Advocates call for developer partnerships with LGBTQ+ groups, stronger safeguards and bans on systems that detect or classify gender.