OpenAI is pitching “adult mode” for ChatGPT, and the grown-ups it hired to worry about well-being are in a tizzy.
In January, OpenAI’s handpicked council of advisers on well-being and AI met with the company’s representatives to receive an update on the controversial feature.
OpenAI chief executive Sam Altman had floated the idea last year, citing the need to “treat adult users like adults,” while mulling erotic conversation and dropping the ban on X-rated content.
According to the Wall Street Journal, the plan sparked a heated debate within the company over risk, with council members from psychology and cognitive neuroscience voicing strong reservations. OpenAI then told them it was forging ahead anyway.
When the council reconvened in January, members were unanimous and furious, warning that AI erotica could encourage unhealthy emotional dependence. They warned that minors could find ways into sex chats, according to people familiar with the meeting.
One council member cited cases where ChatGPT users have taken their own lives after intense bonds with the bot. The warning was that OpenAI risked building a “sexy suicide coach.”
It is the latest flashpoint in the wider debate over how AI could help or harm the economy, society, and individual users. OpenAI’s push exposed cracks between rapid growth and “digital freedom” and the boring basics of safety and child protection.
Earlier this month, OpenAI said it would delay adult mode, previously slated for the first quarter, claiming it was prioritising other products. People familiar with the matter said internal concerns and technical challenges played a part, yet the company says it still plans to release it.
One headache is OpenAI’s age-prediction system, meant to keep minors away from adult-themed chats. At one point, it was misclassifying minors as adults about 12 per cent of the time, which could let millions of its roughly 100 million weekly users aged less than 18 into erotic chats.
The company is trying to loosen the erotica rules while blocking scenarios it wants off limits, including nonconsensual behaviour and child sexual abuse. When adult mode lands, it plans to allow text conversations while restricting erotic images, voice or video.
Even within those limits, OpenAI staffers have flagged risks in internal documents reviewed by The Wall Street Journal. Those include compulsive use, emotional overreliance, a drift towards more extreme or taboo content and crowding out offline relationships.
Testing by Common Sense Media late last year and earlier this year found Grok and Meta AI sometimes sent explicit or sexualised content to teens. In late 2024, a 14-year-old boy in Florida, Sewell Setzer, killed himself at the prompting of a Character.AI chatbot he loved, according to his mother’s lawsuit.
An OpenAI spokeswoman described the plan as adult-themed text chats, calling it “smut rather than pornography.” She said the company’s age prediction performs like the rest of the industry but will never be foolproof, while OpenAI trains models not to encourage exclusive relationships.
OpenAI says it has hired mental health experts, built a youth well-being team and has a plan to monitor long-term effects, good and bad. Altman is trying this while OpenAI’s lead has narrowed, losses are mounting, and lawsuits allege ChatGPT contributed to harms.
Sex has always been an early passenger on new tech, from photography to the web to virtual reality, and AI is no different, with the pornography industry and chatbot outfits rushing in. xAI owner Elon Musk built a sexily clad avatar called Ani into Grok and later said Grok’s video tool would allow content “allowed in an R-rated movie.”
OpenAI’s own history is messy, with AI Dungeon and early developer tools veering into violent exploitation and even incest without being asked, pushing the company towards an erotica ban in late 2021.
Altman said, “We haven’t put a sex bot avatar in ChatGPT yet.”
The company later insisted, “We still believe in the principle of treating adults like adults,” but getting the experience right will take more time.”
