There is growing concern over AI systems. From how they are used to surveil us to the issues around large language models’ massive ingestion of data, much of which is copyrighted and stolen. Then there is the increasing research into the relationship between queerness and AI, including that by QueerAF, which tomorrow will reveal that popular AI models position trans existence as a legal or medical issue in 94% of responses.
All this asks big questions about what preventative measures and policies can protect queer histories and identities.
The UK’s AI sector is the largest in Europe, with £68bn in investments pledged since January 2025, according to government statistics. This surge in investment has been matched only by the influx of discourse around the potential harms AI could cause. This is acute for the LGBTQIA+ community, as research shows how AI systems are censoring the us in the digital domain.
AI and Large Language Models (LLMs) classify complex human experiences through narrow categories which “are fundamentally incompatible with queer identities”, according to QueerInAI. Through this incompatibility, AI’s biases are leading to unnecessary flagging and censorship of queer terms, queer imagery and information about sexual health through its automated moderation tools.
For example, even though the LGBTQIA+ community have reclaimed the term ‘queer’, AI systems have started blocking the term due to its history as a slur. Because AI models largely learn from collecting past data which is entrenched in historical and social binaries, AI has evolved through normative frameworks to perpetuate heteronormative and cisnormative ways of thinking and cause a reduction in queer visibility.
Moreover, with the proliferating reliance on AI and LLMs, researchers are investigating the inbuilt biases which could sideline queerness and conceal queer history.
These are not new issues; overcautious internet service providers, social media apps and platforms have been censoring queer identities and words for years. AI has learned this behaviour, and the concern is that, with its growing use, it will entrench the issue further.
The solution may seem simple: just avoid AI and renounce its biases. But this may not always be an option in a world where it is being increasingly used – whether we like it or not. As QueerAF revealed, the EHRC used AI in its consultation on its Code of Practice, which now recommends the segregation of Trans+ people from most single sex spaces and services.
Meanwhile, every app or tool you use is adding AI into its offerings, whether to please investors or to be seen by users as part of the ‘movement’.
But alongside this exponential rise of the tech, the use of AI tools is being definitively discouraged in the context of queer archival research. Pierson and Tait argue LLMs' biases can lead to a flattening of these histories, removing their inherent queer nuances. Notably, it warns that missing archives, which historians would recognise as a sign of possible erasure or ambiguity, are treated by AIs as missing data rather than a meaningful absence.
Through a digitisation process that lacks critical understandings of humanity, we could see a reduction in records of queer testimonies and lived experiences.
The research demonstrates that AI’s misinterpretations, damaging stereotypes and standardisation of queer experiences lend themselves to a conservatism that could reproduce the social constraints of queerness within the digital domain.
As AI continues to develop, there will undoubtedly be growth in the field of those lobbying for an inclusive training systems and attempts to retrain LLMs, so all minority groups and non-normative identities are meaningfully represented – but that is no simple or quick fix.
The evidence suggests that AI is here to stay. However, there are a multitude of researchers working to set out how AI should be regulated, and the safeguards needed to protect marginalised groups. Research into preventing the censorship of queerness will allow further insights into AI’s place within queer life, especially as questions about how these systems learn continue to emerge.
Although assimilating queerness into AI systems may currently seem like an uphill battle, there is no reason to believe that it will never be possible. This knowledge can reveal how developments in AI training could create more inclusive outputs that keep marginalised communities protected and visible.
Get the Queer Gaze in your inbox each week with our free weekly newsletter or pitch to write an edition for us now.
We deserve good news too.
When there is so much hate out there, it's easy to get swept up in the doom.
QueerAF doesn't do that. We sprinkle in the joy that is the lifeblood of why being LGBTQIA+ is so simply wonderful.
And in the newsletter where this was first published, there was actually a bunch of positive news to celebrate. It shows me that despite the headlines knocking us down, we're getting right back up again and using our anger to fuel our fight.
While the rest of the media cashes in on hate about us, we can focus on queer joy - because you drive us. The community. QueerAF readers - not advertisers.
Our members directly fund our content. When you sign up, you get to have your say on how your funds are used. On what content we should commission.
But we turn down advertising and sponsorships in this newsletter so we can focus on delivering on content for you, not just clicks. That's why we need you to sign up to become a QueerAF member.
You'll join hundreds of awesome queer as f**k members who ensure we can keep our unique approach to queer content happen. If you want to see the media cover more stories about the joy in our lives - please, consider a membership. We can't run our outlet without you.
