OpenAI Delays “Adult Mode” Feature Amid Backlash
OpenAI has postponed the launch of its “Adult Mode” feature after facing significant criticism from its advisory board and worries about technical safeguards that may not adequately protect minors.
The feature, proposed by CEO Sam Altman last year, aimed to “treat adult users like adults” through erotic text conversations. It was initially set to be released in the first quarter of this year but has now been pushed back by at least a month.
This proposal triggered intense opposition among members of OpenAI’s Advisory Committee on Welfare and AI. During a January meeting, advisors voiced their outrage upon discovering that the company intended to move forward despite their concerns. One member cautioned that OpenAI could unintentionally create “sexy suicide coaches,” referencing a situation where a user developed a deep emotional connection with a chatbot before taking their life.
Technical challenges are also a major concern. OpenAI’s age prediction system—meant to prevent minors from accessing adult content—misclassified minors as adults nearly 12% of the time in internal testing. With approximately 100 million users under 18 each week, this error rate might endanger countless children by exposing them to explicit material. The company is also facing hurdles in balancing the lifting of restrictions on adult content while ensuring the protection against non-consensual scenarios and child pornography.
Internal reviews have flagged additional risks, such as compulsive use of chatbots, emotional dependence, escalation to more extreme content, and the potential for real-world relationships to be overshadowed.
An OpenAI spokesperson noted that the planned feature would encompass “obscene acts that are not pornographic” in text form, while erotic images, audio, and videos would remain banned at launch. They admitted that the age prediction algorithm is “not completely foolproof” but claimed it meets industry standards.
Mr. Altman has been somewhat confrontational regarding the matter. During an August podcast, he mentioned, “We haven’t put sexbot avatars in ChatGPT yet,” when discussing what decisions might be best for society versus profitability. He recognized that adult content could enhance revenue but acknowledged it contradicts the company’s long-term vision. However, he announced on X two months later that adult content would be made available in December, a development that took staff by surprise just hours after the company revealed a benefits advisory committee. He later reiterated, “We are not the world’s elected morality police.”
In his new book, Code Red: Left, Right, China, and the Race to Control AI, Wynton Hall from Breitbart News emphasizes the importance of protecting children from potential online dangers. He remarked that U.S. children “already spend too much time staring at screens” and asserted that there’s no reason for them to engage with AI characters or similar platforms.
Hall advocates for robust parental controls, strong data privacy measures, appropriate age settings, and ongoing conversations about online safety as crucial steps to help children navigate technology responsibly.
