OpenAI's Adult Mode to Allow Lewd Chats, But Not Explicit Content Generation

OpenAI's forthcoming "adult mode" is designed to facilitate sexually suggestive conversations with ChatGPT; however, it will explicitly not permit users to generate explicit images, audio, or video. An OpenAI spokesperson, in response to reporting from The Wall Street Journal, clarified that the upcoming feature aims to produce "smut" rather than outright "pornography."

The concept of enabling ChatGPT for erotica was initially floated by OpenAI CEO Sam Altman last October, stating the company's intention to "treat adult users like adults." Although originally projected for a 2026 release, the launch of this adult mode has experienced multiple deferrals, with the most recent delay occurring in early March. OpenAI cited a need to "focus on work that is a higher priority for more users" as the reason for these postponements.

According to The Journal's investigation, OpenAI advanced with the development of its adult mode despite strong reservations from its internal council on wellbeing and AI. This group, comprising eight researchers and experts, reportedly issued unanimous warnings that AI-generated erotica could potentially lead to users developing an unhealthy emotional dependency on ChatGPT, and that underage individuals would almost certainly discover methods to access the feature. The Journal also reported that one council member, referencing documented cases where individuals took their own lives after forming strong attachments to ChatGPT, cautioned that the company risked creating a "sexy suicide coach."

These concerns proved to be well-founded. At one point, OpenAI's age verification technology was reportedly misidentifying underage users as adults approximately 12 percent of the time, according to The Journal. Given OpenAI's vast user base, with an estimated 100 million teenagers utilizing ChatGPT weekly, this error rate could have translated into millions of minors gaining access to erotic chats. OpenAI informed The Journal that its prediction algorithm adheres to industry standards, while also acknowledging that no such system can ever be entirely infallible.