OpenAI was originally scheduled to launch this quarterChatGPT "Adult Mode"The launch will be delayed due to difficulties in defining content review standards and deep concerns about the protection of minors.The Wall Street Journal reportedThis controversial feature will initially only support text conversations with adult themes, without the ability to generate images, audio, or video. An OpenAI spokesperson described the content allowed by this mode as "smut" rather than "pornography," attempting to draw a blurred line between freedom of speech and content control.
From "serious mental health issues" to "verified adult pornography"
The idea for this feature dates back to last October. OpenAI CEO Sam Altman stated at the time that the company had successfully mitigated the "serious mental health issues" caused by its AI models, thus enabling it to relax safety restrictions and introduce erotic content for "verified adult users." This announcement sparked widespread discussion in the industry and was seen as a major shift for OpenAI regarding the ethical boundaries of AI.
Earlier this month, OpenAI abruptly announced a delay in the feature's release, citing the need to prioritize higher-priority tasks. A new release schedule has yet to be announced. However, according to internal sources obtained by the Wall Street Journal, the real reasons for the delay are far more complex than a "priority adjustment," involving deep internal concerns about content moderation mechanisms and the technical challenges of protecting minors from inappropriate content.
Internal controversy: From "sexy suicide coach" to age verification failure
In January, an internal advisory committee at OpenAI issued a stern warning to the company, pointing out that ChatGPT's adult mode could be exposed to children and could even foster unhealthy emotional dependence on the chatbot. One committee member, who wished to remain anonymous, went further, using a sensational but vivid metaphor: OpenAI might be creating a "sexy suicide coach."
Content moderation is the core issue causing the delay. Sources familiar with the matter revealed that OpenAI is working to remove ChatGPT's restrictions on NSFW (Not Suitable for the Workplace) content, while simultaneously ensuring that more harmful scenarios—such as non-consensual acts or child sexual abuse—remain strictly prohibited, making this line of demarcation far more difficult than anticipated.
The most worrying aspect is the accuracy of the age prediction system. OpenAI developed this system to prevent minors from accessing pornographic content, but during the testing phase, it reportedly misclassified minors as adults at a rate of approximately 12%. Considering that ChatGPT attracts approximately 1 million users under the age of 18 each week, such an error rate means that millions of minors may have had the opportunity to engage in sexually suggestive conversations with the chatbot.
An OpenAI spokesperson stated that the company's age prediction algorithm performs comparably to other companies in the industry, but "it can never be 100% perfect."
Text vs. Images: Strategic Choices for Overcoming Regulatory Loopholes
It's worth noting that OpenAI's initial decision to limit its adult mode to text-based conversations is not merely a technical limitation, but more likely a well-thought-out regulatory circumvention strategy. The UK's Online Safety Act requires online platforms to implement age verification for pornographic images, but it does not have the same stringent requirements for text-based erotic content, giving OpenAI's text-only strategy a clear regulatory advantage.
This strategy contrasts sharply with its competitors. xAI's Grok chatbot recently announced that its image and video generator will allow any "content permitted in R-rated movies," with CEO Elon Musk describing Grok's companionship feature as "spicy." In comparison, OpenAI's text-based restriction strategy appears more conservative and is more likely to operate within the existing regulatory framework.
OpenAI's Dilemma: Freedom of Speech vs. Social Responsibility
OpenAI's hesitation regarding its adult mode reflects a fundamental dilemma facing the generative AI industry: how to strike a balance between openness and security. On the one hand, as AI models become increasingly capable, users have a clear demand for a freer and more unrestricted dialogue experience; on the other hand, as a platform with hundreds of millions of users, OpenAI must shoulder the social responsibility of preventing minors from accessing inappropriate content and avoiding reinforcing users' negative psychological states.
Of particular concern is the advisory committee's warning about "emotional dependence." When AI chatbots can not only provide information but also act as emotional companions or even sexually suggestive roles, their potential impact on users' mental health should not be underestimated.
