ChatGPT’s much-anticipated (and somewhat delayed) adult mode will be able to create “smut rather than pornography”, a spokesman for the AI chatbot’s company OpenAI has said.
The Western world’s most prominent AI chatbot, launched in November 2022, has so far been programmed to resist being drawn into highly-sexualized or NSFW conversations. However, OpenAI has been working on an age-restricted version that will be able to engage in racy chat.
Now an OpenAI spokesperson has told the [paywalled link] Wall Street Journal (WSJ) that this chat is likely to be “smut”-level rather than anything properly pornographic. The spokesperson also reiterated that the new adult ChatGPT mode will not be able to generate audio, images or video, so will be a text-only service, just with fewer guardrails.
The erotica question
ChatGPT has so far been firm in its refusal to engage in purely titillating sex talk, and although some ‘jailbreak’ hacks to the AI chatbot have periodically made it break its guardrail rules, these have usually been quickly shut down.
In early 2025 ChatGPT started allowing some erotic chat, but only in contexts such as being educational or medical.
So it came as a surprise to many when, in 2025, it was revealed that an “adult mode” was in the works. The mode will allow verified adults to engage in erotic chat with the chatbot, with OpenAI CEO Sam Altman saying that the move was part of a shift towards “treating adult users like adults”.
As well as telling the WSJ that the new adult mode would be able to make “smut rather than pornography”, the OpenAI spokesperson said that the chatbot would encourage users to seek relationships (smutty or otherwise, presumably) in the real world.
Age verification, unresolved
OpenAI has framed the launch as contingent on robust safety infrastructure, which is the kind of claim worth scrutinizing. The company’s age prediction system, according to unnamed insiders in the WSJ report, misclassified children as adults in roughly 12 percent of cases, which, if accurate, is a figure that sits uneasily alongside any assurance that safety is in hand.
The chatbot sector is under intense scrutiny. Character.AI has faced a lawsuit from the family of a Florida 14-year-old who died by suicide after reportedly intensive chatbot use, and has since moved to restrict open-ended conversations for under-18s, a case we covered in detail as part of our look at how tech giants are splitting on erotic AI. In the US, bills have been proposed that would force companies to limit chatbots’ addictiveness.
The new ChatGPT adult mode will use age prediction as a way of verifying a user’s age, which in theory would prevent anyone under 18 from being exposed to “smut” talk. However, the WSJ has reported that there are concerns in the company about the age estimation’s effectiveness so far. Unnamed sources within OpenAI reportedly said that its age prediction system was misclassifying children as adults around 12 percent of the time.
Earlier in March 2026, OpenAI said that the launch of ChatGPT’s adult mode was being delayed, and it may be the case that the company smooths out these issues in time for the launch, whenever that might be.
Internal dissent, and a firing
The forthcoming launch remains controversial, both in and out of the confines of OpenAI’s offices. The company recently fired Ryan Beiermeister, who had worked on product policy for the company, and who was reportedly outspoken about concerns about ChatGPT delving into erotica.
OpenAI said that Beiermeister was sacked for sexual discrimination against a male colleague, rather than anything to do with opposition to ChatGPT’s adult mode. Beiermeister said the claims of sexual discrimination were “absolutely false”.
Will ChatGPT’s adult mode be a game-changer for AI chatbots when it is finally released? It’s certainly significant that a massively mainstream chatbot is nudging towards more NSFW content, after years of resisting erotica, and often being accused of being far too conservative when shutting down chat requests.
That age estimation technology better be rock-solid by then, though, or OpenAI’s catalog of controversies is quickly going to get a new chapter.
Leave a Reply