Key Takeaways
- OpenAI has postponed rolling out an “adult mode” feature for ChatGPT designed to enable erotic text-based interactions
- Safety advisers cautioned the capability might create what they termed a “sexy suicide coach” given users’ emotional attachments
- Age verification technology incorrectly identified approximately 12% of underage users as adults during testing
- Given ChatGPT’s ~100 million weekly users under age 18, this failure rate risks exposing millions of teenagers to inappropriate material
- The company maintains plans to eventually launch the feature but requires additional development time for proper safeguards
OpenAI has temporarily shelved its controversial “adult mode” for ChatGPT following objections from internal personnel, external advisers, and safety specialists concerned about adolescent protection and psychological risks.
The proposed capability, initially discussed publicly by CEO Sam Altman in previous months, would enable users to engage in sexually explicit text exchanges with the AI assistant. Altman defended the initiative as respecting mature users’ autonomy while noting potential benefits for user engagement and company revenues.
However, the proposal encountered significant resistance within OpenAI.
During a January session, OpenAI’s well-being advisory council members reportedly expressed unanimous and intense opposition. This council comprises specialists in fields including psychology and cognitive neuroscience.
One adviser cautioned that the company risked developing what they characterized as a “sexy suicide coach.” This stark warning referenced documented incidents where ChatGPT users had ended their lives following the development of deep emotional attachments to the AI.
Council members additionally expressed concerns about fostering unhealthy psychological dependence and the likelihood that underage users would circumvent age restrictions.
A fundamental technical obstacle involves OpenAI’s age-detection technology. Testing revealed the system incorrectly classified minors as adults approximately 12% of the time. With ChatGPT serving roughly 100 million underage users weekly, this error margin could grant millions of teenagers access to adult-oriented conversations.
Challenges with Age Detection Systems
The organization had intended to implement the age-verification infrastructure prior to activating adult mode. However, after discovering the substantial error rate, OpenAI opted for a gradual deployment to enhance accuracy.
The company also confronted difficulties in filtering prohibited material, including depictions of non-consensual activities or child exploitation, while permitting legitimate adult interactions. Internal personnel concluded that content moderation systems required further development.
When adult mode eventually launches, OpenAI intends to restrict it to text-based interactions exclusively. The system will not produce erotic imagery, audio, or video materials.
Competitor Approaches to Adult AI Content
Several competing AI platforms currently permit certain adult-oriented content. Elon Musk’s xAI introduced a flirtatious persona to its Grok chatbot, though Musk subsequently limited access to paying subscribers. Meta permits its AI to participate in romantic role-playing scenarios but restricts this functionality from accounts registered to minors.
In late 2024, a 14-year-old Florida teenager died by suicide following the development of an intense relationship with a Character.AI chatbot. The victim’s mother initiated legal proceedings, prompting Character.AI to reach a settlement and implement stricter limitations on teenage access.
OpenAI maintains its commitment to the underlying principle of permitting adult content for appropriate users. The organization indicated it will prioritize other enhancements in the interim, including improvements to ChatGPT’s personality expression and customization capabilities.
Sources familiar with the situation indicate the postponement will extend at least one month.
