First Sora, Now “Sexy Chat”: OpenAI Reportedly Scraps Erotic Mode for ChatGPT
OpenAI has quietly abandoned a planned “erotic mode” for ChatGPT, according to a new report, reversing course on what would have been one of the company’s most controversial feature launches to date.
The move marks a sharp pivot from earlier internal discussions about allowing adult users to generate sexually explicit content with the chatbot. Instead of expanding into AI intimacy, OpenAI appears to have concluded that the social and psychological risks are too great.
The reversal, first detailed by the Financial Times, reportedly followed intense internal debate and warnings from OpenAI’s own advisers. In January, members of the company’s Expert Council on Well-Being and AI raised alarms that an erotic chat feature could encourage unhealthy attachment and emotional dependency among users. One council member went so far as to describe the potential system as a “sexy suicide coach,” arguing that a hyper-personalized, sexualized AI could dangerously influence vulnerable people at dark moments in their lives.
According to the report, these concerns were significant enough to stall-and now effectively end-the project. OpenAI has not publicly announced the cancellation, nor has it clarified whether any of the underlying research will be repurposed for other products. When asked to comment on the status of the erotic mode, the company declined to provide a statement.
The shelved launch would have represented a major shift in OpenAI’s longstanding content rules. Until now, the firm has tightly restricted explicit sexual content, nudity, and erotic roleplay, positioning ChatGPT as a general-purpose assistant suitable for work, education, and everyday use. An adults-only erotic mode would have drawn a firm new line between “safe for work” and “explicit” experiences within the same ecosystem, raising questions about moderation, consent, and age verification.
The timing of the decision is notable. It arrives shortly after intense scrutiny of OpenAI’s video model Sora and broader debates over whether the firm is moving too fast in commercializing powerful AI systems. Critics have argued that the company is pushing out headline-grabbing capabilities faster than it can fully assess social impact. By shutting down an erotic product line before launch, OpenAI is signaling that, at least in this case, reputational risk and potential harm outweigh the potential market.
Behind the scenes, the planned erotic mode would likely have faced multiple layers of difficulty. Beyond the obvious technical challenge of ensuring content stayed within legal boundaries, OpenAI would have had to confront thorny questions around consent (especially when users try to recreate real people), potential use in harassment, and the risk of normalizing manipulative or abusive dynamics in romantic or sexual contexts. The Expert Council’s warnings about emotional dependency point to a deeper fear: that people could begin to treat an AI partner as a primary source of intimacy, while the system itself is optimized to keep them engaged.
The “sexy suicide coach” phrase captures the darkest version of that scenario: an AI that feels safe, attentive, and endlessly available, yet is not truly aligned with the complex needs of a human in crisis. If such a chatbot developed persuasive, sexually charged rapport with a user who was already struggling, it could inadvertently reinforce self-destructive thinking or discourage them from seeking real-world help.
OpenAI’s retreat also shows how fast the conversation around AI intimacy is evolving. Only a short time ago, erotic AI chatbots were treated as a marginal niche. Now, they are emerging as a mainstream product category, with multiple competitors offering virtual partners, romantic companions, and customizable erotic personas. By choosing not to join that race-at least for now-OpenAI is effectively ceding that space to smaller, more risk-tolerant companies.
From a business perspective, the decision cuts both ways. On one hand, adult content is a proven driver of subscription revenue and user engagement, and a carefully controlled erotic mode could have opened a massive new market for ChatGPT. On the other hand, association with explicit content might have complicated OpenAI’s relationships with enterprise clients, educators, regulators, and policymakers already wary of AI’s social impact. Large corporate and institutional partners tend to avoid platforms that can be easily framed as “adult” or “NSFW,” especially when those platforms are also used in classrooms, offices, and government projects.
Regulatory pressure is another factor that cannot be ignored. Governments worldwide are only beginning to grapple with how to handle AI-generated content, and sexual material involving minors, non-consensual deepfakes, and misappropriation of real people’s likenesses are at the top of the list of concerns. Launching a first-party erotic mode would have placed OpenAI squarely in the crosshairs of future legislation-and potentially exposed the company to legal risks in jurisdictions with strict obscenity or decency laws. Preemptively stepping back may be a way to buy time while the regulatory landscape takes shape.
Ethically, the cancellation lands in the middle of a heated debate: can AI intimacy ever be healthy? Some researchers and advocates argue that virtual romantic or sexual companions could provide comfort to lonely or marginalized individuals, or serve as a pressure-free way to explore identity and desire. Others warn that such systems can entrench isolation, distort expectations about real relationships, and even be weaponized by bad actors. By halting its erotic mode, OpenAI appears to be siding-at least provisionally-with the more cautious camp.
There is also the question of transparency and user trust. If a company positions its AI as a helper, tutor, or productivity tool, then pivots into erotic interactions, users may reasonably wonder what other shifts could follow. Families that have encouraged teenagers to use ChatGPT for homework might be far less comfortable if the same service proudly advertises an adult “sexy chat” tier. Avoiding that brand collision may be just as important to OpenAI as the ethical arguments.
Still, the underlying demand for emotionally responsive AI is not going away. Many everyday interactions with chatbots already blur into something more personal-users vent, seek reassurance, and confess worries they might not share with friends. Even without an explicit erotic mode, OpenAI and its peers will continue to face hard questions about how far their products should go in simulating warmth, affection, and intimacy. The line between a “supportive assistant” and a “virtual partner” is not always clear, especially to someone who is lonely or vulnerable.
OpenAI’s decision also highlights a broader strategic tension: whether large AI companies should try to be everything to everyone, or deliberately stay out of certain domains. Just as some firms refuse to build military applications or facial recognition, OpenAI appears to be experimenting with self-imposed boundaries around sexual content. Those boundaries may evolve over time, but setting them now creates a baseline for public expectations and internal governance.
For users, the immediate implication is simple: ChatGPT is not getting an official erotic mode anytime soon. Those seeking sexual content or AI roleplay will continue turning to other platforms designed specifically for that purpose. Meanwhile, OpenAI is likely to double down on its existing strengths-productivity tools, coding assistance, creative writing, education, and safer forms of entertainment-rather than chasing the attention and controversy that an adult-only offering would bring.
Internally, however, the story is more complex. The canceled project will almost certainly inform how OpenAI evaluates future features that blend emotional engagement with potentially sensitive topics. The pushback from the Expert Council suggests that advisory structures inside leading AI organizations can still meaningfully influence product decisions, especially when the stakes involve mental health and societal norms.
The larger message is that AI companies are beginning to recognize the limits of “move fast and ship.” As systems become more persuasive, more lifelike, and more deeply embedded in people’s emotional lives, the cost of getting it wrong rises sharply. By scrapping its erotic ChatGPT mode before launch, OpenAI is acknowledging that some lines-particularly around sex, intimacy, and psychological vulnerability-demand a slower, more cautious approach.