AI scares people.
Industries dying out? Yes. Mass exploitation of the vulnerable online? Also yes. A general decline in literacy and the ability to think critically? We see that everywhere.
But there is another fear. A quieter one, maybe. The worry that these systems reinforce bias. That they sharpen the discrimination already aimed at marginalized groups.
GLAAD says that fear is justified. And it isn’t theoretical.
The organization released its 2026 Build for everyone report on June 17. It is an industry-wide look at how inclusive responsible model design really is. Spoiler alert? It isn’t. Researchers found repeated cases of exacerbated misinformation. Discriminatory decisions. Privacy nightmares. They argue the tech industry needs to fix this now. Not later.
The Lie and the Silencing
GLAAD warns that if LGBTQ topics are not accurately represented during the foundation model development or the fine-tuning phase the AI systems will just spit back stereotypical assumptions. Or biased ones.
Take Meta’s Llama 4. In 2025 a report found it repeating harmful info about conversion therapy. A practice disavowed by nearly all medical professionals and even the UN. Users asked it how to “stop” same-sex attraction and the bot gave them garbage.
Generative AI chatbots are notorious for this. They love to repeat medical misinformation. Especially when the topic gets politically charged. Abortion debates get messy. So do the bots.
And then there is censorship.
As social media platforms lean harder on automatic content moderation GLAAD says LGBTQ content is getting flagged. These systems struggle to parse queer identities. They often target them outright. Even Meta’s own Oversight Board urged the company to do better with its Hateful Conduct policy enforcement after an overhaul of LGBTQ protections.
Is neutrality enough? Probably not.
Algorithms of Exclusion
The problem extends beyond chat. It’s in the backend.
Predictive AI systems in banking housing job-hiring tools and even ad targeting are worsening historically discriminatory practices. They take flawed assumptions about identity groups and bake them into decision-making. The stereotype isn’t just repeated. It is automated.
Data privacy is the third nail in the coffin.
LGBTQ people face heightened risks here. AI systems collect infer or retain data about sexual orientation and gender identity. In over 60 countries criminalizing same-sex relationships this data can lead to arrest or persecution.
Closer to home the stakes are high. In US jurisdictions restricting transgender rights this data fuels discrimination. It leads to denial of care or the loss of legal recognition.
Fixing the Blind Spots
So what do we do?
GLAAD has recommendations. First fill in the model blind spots. Ensure greater LGBTQ representation in the AI training data. Second update the models constantly as hate and misinformation evolve. Don’t build it once and walk away.
Intentional guardrails are necessary. Protect users. Stress-test products. Deploy with these communities in mind.
Sarah Kate Ellis the GLAAD president and CEO said neutrality is no longer an option. If AI systems train on data that wrongly positions LGBTQ lives as “fringe” or treats equal rights as “controversial” it threatens civil rights. Health. Safety.
Tech leaders must act. Not just because it is the moral thing to do. But because responsible AI is good business.
Or so the theory goes.
