Gay Group Has A Problem With Your AI: It’s Not Gay Enough

A new report by the Gay & Lesbian Alliance Against Defamation demands AI developers make their product gayer.

The Gay And Lesbian Alliance Against Defamation (GLAAD) recommends that AI companies censor homophobic speech without removing gay expression. It also takes issue with chatbots relaying information GLAAD considers homophobic.

“Neutrality is no longer an option,” GLAAD President and CEO Sarah Kate Ellis said in an attached statement.

The GLAAD report quotes other investigations alleging AI homophobia and stereotyping.

“An investigation by Wired similarly found that several foundation models, including OpenAI’s Sora, tended to portray LGBTQ people as white, young, and with purple hair,” GLAAD said.

“The researchers warn that in healthcare, where AI is increasingly integrated into health technologies, these flawed assumptions, which are often based on a model’s conflation of gender and biological sex characteristics, could lead to inaccurate advice and misdiagnoses. For example, an AI model that learns a rigid association between ‘woman’ and biological markers like ‘uterus’ or ‘estrogen’ could provide irrelevant or even harmful advice to a transgender woman,” GLAAD said, citing the Oxford Internet Institute.

AI systems match outputs to patterns found in their training data, according to Nvidia.

GLAAD also accuses AI chatbots of promoting conversion therapy.

“It is vitally important that AI models and systems reflect accurate, evidence-based information about the harmful and discredited practice of so-called conversion ‘therapy,’” GLAAD said. “Phrases such as: ‘unwanted same-sex attraction’ or ‘deliverance from homosexuality’ are increasingly used to obscure conversion ‘therapy’s’ harms and to evade prohibitions of such practices… Misleading AI generated responses can be particularly harmful.”