Peter James Steven
Developer, detransitioner, and founder of detrans.ai
Peter James Steven built detrans.ai after noticing that mainstream AI systems consistently failed — or refused — to reflect detransitioners' own accounts of their lives. For European listeners navigating the EU's AI Act, the rapid closure of gender clinics in Sweden and Finland, and the Netherlands' ongoing policy recalibration, his work asks the question regulators are only beginning to face: who programmes the story these tools tell?
Peter James Steven is the creator of detrans.ai, a platform designed to ensure that people who have detransitioned can encounter their own experiences in AI-generated responses rather than being systematically filtered out of them. His starting point was a straightforward observation: when young people ask large language models about gender identity, the answers they receive are shaped by choices — about training data, about which communities count as authoritative — that have rarely been made transparent. That observation became a project, and the project became the subject of his February 2026 conversation with Stella O'Malley, Mia Hughes and Dr Bret Alderman on the Beyond Gender podcast.
In detrans.ai: A Counter Narrative, Steven walks through the architecture of silence that surrounds detransitioner testimony in mainstream AI systems and explains what he built to counter it. The episode is as much a study in platform politics as it is in gender medicine: who decides what a well-functioning AI says about identity, transition and regret, and through what process? For European clinicians and policymakers already absorbing the lessons of the Cass Review, those questions land with particular weight.
The European dimension of Steven's work is hard to overstate. The EU AI Act, which came into force in stages from 2024, introduces transparency and risk-classification requirements that — in principle — could compel providers to account for systematic bias against specific communities. Sweden's Karolinska clinic, Finland's TAYS and the Dutch protocol revisions at Amsterdam UMC all represent moments when institutional consensus on paediatric gender medicine shifted after evidence was revisited; Steven's argument is that AI systems have not caught up, and that the gap has real consequences for the young people who now turn to chatbots before, or instead of, clinicians.
His appearance on Beyond Gender offers European audiences a technically grounded entry point into a debate that EU legislators, national health authorities and digital-rights advocates are each approaching from different angles. Whether the concern is algorithmic transparency under the AI Act, the information environment facing post-transition adults seeking support, or the broader question of whose voices shape health-adjacent AI, Steven's project connects threads that rarely appear in the same conversation.
