Developing Human Sexuality in the Age of AI: Information, Consent, and Responsibility

Human sexuality has always been shaped by technology, from the printing press and Victorian photography to the internet. Generative AI is the latest development, and its effects on how people learn about and engage in sexuality deserve careful analysis. This article focuses on two areas: the distribution of sexual information and the principle of consent.

Generative AI uses machine learning models trained on vast amounts of human-created text, images, video, and audio. Because the internet's content includes sexual material, models inevitably absorb exposure to sexuality. Guardrails, reinforcement learning, and prompt engineering can limit outputs, but generative AI is broadly as capable of producing sexual content as any other content. The model has no inherent sense of explicitness; meaning and social norms come entirely from the humans who shape its training.

Researcher Nicola Döring and colleagues identified four main ways users interact with AI in sexual contexts: sexual information and education, sexual counseling and therapy, sexual and romantic relationships, and erotica and pornography. Mainstream providers generally restrict explicit interactions, but such restrictions cannot be universally enforced, and demand for sexual content is strong, so the market is likely to supply it in some form.

Many people already seek sexual health information online because it is convenient, anonymous, and free from judgment. LLMs extend this trend. Studies suggest the accuracy of basic sexual health information from mainstream LLMs is not worse than general web searches. However, LLMs present answers in an authoritative tone and often obscure their sources, making AI literacy essential. Hallucinations remain an unavoidable problem, so skepticism and verification are important for users.

LLM perspectives reflect the choices of their developers. Human raters in reinforcement learning bring their own beliefs, and on matters of opinion, companies effectively decide what is acceptable. Research has found that LLM responses can minimize or devalue non-mainstream sexual expression, including LGBTQ+ perspectives. Refusal, where models decline sexual topics entirely, leaves users without information and can imply that sexuality is shameful. Companies face difficult questions about whose moral standards to follow and where to draw boundaries.

Learning about sexuality also comes from observing images and video. Studies show image generators carry implicit bias: asked to depict a couple, one widely used generator produced a young, able-bodied, white, mixed-sex couple with a more sexualized woman. Users who do not know what they want may absorb these biases. Generative AI also produces photorealistic images of physically impossible bodies, escalating the unrealistic beauty standards once fueled by airbrushing and pornography, and distorting expectations for both boys and girls.

Consent means that all parties readily and affirmatively agree throughout, without coercion or manipulation, and this applies to creating and sharing sexual imagery as well as physical interaction. Generative AI raises pressing questions: how do we know whether a person consented to their likeness being used sexually, what rights they hold, what obligations AI companies and users bear, and how easy access to AI-generated content affects users' understanding of consent when they never interact with the people depicted. Deepfakes predate generative AI, but cheap, accessible tools and declining detectability have changed the scale of the problem.

Questions of ownership over one's likeness parallel copyright debates faced by authors, actors, and artists. Generative AI products place few effective restrictions on depicting public figures sexually, and safeguards are easily circumvented. Sex workers and erotica or pornography creators face the same risk: consenting to create sexual content does not mean consenting to having their work or image used to train or generate AI content. Social stigma should not weaken these rights, and the legal precedents set now may affect many other scenarios.

Scholars increasingly use the term AI-generated image-based sexual abuse rather than revenge porn, since it better captures the harm of sexual content made from someone's likeness without permission. Victims can be deeply traumatized regardless of whether viewers believe the content is real, and the severity of harm does not depend on the victim's public profile. Although image manipulation has long existed, generative AI makes abuse easier, cheaper, and more accessible. OpenAI's planned move into erotica raises concerns, since its safety discussion focused on user well-being rather than the rights of nonconsenting individuals whose likenesses could be used.

Generative AI is changing how people learn about sexuality and how they understand consent. Its benefits, such as anonymous access to accurate health information, must be weighed against risks including biased standards, distorted body images, and the normalization of using others' likenesses without permission. Young people, who learn implicitly that someone's likeness is fair game when technology is involved, are particularly affected. Protecting individual safety and rights while preserving freedom of sexual expression requires understanding the current landscape and addressing consent, accountability, and transparency before these products expand further.