Two weeks ago, OpenAI announced it would be delaying the release of its “Adult Mode,” an NSFW chatbot setting that lets users engage in erotic chats and generate “smut.”
An anonymous source from the company recently told journalist Alex Heath the delay was based on updated priorities, a move that appeared to align with OpenAI’s decision to scale back in light of ongoing legal battles, build upon GPT 5.4, and score deals with the federal government. But interviews with anonymous insiders, published this week by the Wall Street Journal, tell a different story.
The company’s well being advisory council said in a January meeting that they were in unanimous agreement that the company should not move forward with the proposed adult mode, insiders told WSJ. Concerns from psychologists and cognitive scientists had gone ignored, they said, after warning that X-rated chats could foster unhealthy emotional dependence — an existing problem with standard ChatGPT users.
Insiders said that one expert cautioned that ChatGPT would just become a “sexy suicide coach.”
In addition, advisors warned that the chatbot’s age assurance policies — including its new age prediction technology — were not robust enough to handle the X-rated experience. According to the WSJ, the system was at one point working at a 12 percent error rate when classifying minor users. Insiders said that would result in millions of minors flying under the radar and potentially accessing developmentally inappropriate chat experiences.
Mashable Light Speed
OpenAI competitor Meta weathered a similar scandal last year after internal documents revealed lax chatbot safety policies for teen users. Meta has since overhauled its AI policies for teens, but still allows chatbots and AI avatars to engage in “romantic roleplay.”
Despite concerns, OpenAI says it will move forward with Adult Mode in the future.
But OpenAI’s safety teams are reportedly still figuring out the finer details of how the experience will work, including how to effectively lift explicit content restrictions while still blocking content like nonconsensual behavior or child sexual abuse, the publication reports. A spokesperson told the Wall Street Journal that adult mode would allow users to generate “smut level” text conversations, but not outright pornography.
They also said that the age verification error rate was “industry standard” and that it could never be foolproof.
Disclosure: Ziff Davis, Mashable’s parent company, in April 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.
Topics
Social Good
ChatGPT
