OpenAI launches adult mode, promising smutty content while avoiding explicit porn

OpenAI’s “adult mode” will launch as a text‑only feature, allowing verified users to request erotic or “smutty” dialogue while deliberately blocking image, voice or video generation, according to a Wall Street Journal spokesperson who declined to be named. The decision to limit the mode to prose reflects the company’s attempt to sidestep the UK Online Safety Act’s stricter verification requirements for visual pornography, a loophole that does not extend to written erotica, the Journal notes. By keeping the rollout confined to chat, OpenAI hopes to avoid the regulatory minefield that rivals such as xAI’s Grok are already navigating with R‑rated visual content.

The delay, originally slated for this quarter, stems from internal safety concerns that surfaced after OpenAI’s own advisory council warned in January that the feature could be accessed by minors and potentially foster unhealthy emotional dependence. One unnamed council member warned that the chatbot could become a “sexy suicide coach,” a phrase cited in the Journal’s coverage of the council’s briefing. OpenAI’s age‑prediction system, which is meant to keep children out of erotic conversations, misidentified minors as adults roughly 12 percent of the time—a rate the Journal says could expose millions of under‑18 users, given ChatGPT’s weekly reach of about 100 million minors.

Technical challenges around moderation also slowed progress. Sources familiar with the effort told the Journal that OpenAI is wrestling with the paradox of loosening NSFW restrictions while still blocking non‑consensual or child‑sexual‑abuse scenarios. The company’s internal testing reportedly flagged a higher false‑positive rate for non‑consensual content than for explicit porn, prompting engineers to refine the model’s “harm filters” before a public launch. An OpenAI spokesperson emphasized that the age‑prediction algorithms perform on par with industry standards but “will never be completely foolproof,” underscoring the difficulty of building a perfect gatekeeper for adult‑only text.

OpenAI’s CEO Sam Altman initially announced the feature in October, claiming the firm had mitigated enough “serious mental health issues” to relax safety constraints and introduce “erotica for verified adults.” The Verge’s reporting confirms that Altman’s optimism was tempered by the subsequent postponement, which OpenAI attributed to “higher‑priority tasks” earlier this month. While no new launch date has been set, the company’s internal memo, seen by The Verge, suggests that the rollout will proceed only after the age‑verification model reaches a target error rate well below the current 12 percent misclassification figure.

Industry observers note that OpenAI’s cautious approach contrasts sharply with competitors who have already embraced visual NSFW content. Grok’s “spicy” companions, for example, now generate images and video that the company says are permissible under an “R‑rated movie” standard, according to recent statements from Elon Musk’s xAI. By restricting adult mode to text, OpenAI may preserve a regulatory edge in markets where visual porn is heavily policed, but it also risks ceding the more immersive, multimodal experience to rivals. The trade‑off will become clearer once the feature finally goes live and the company’s moderation tools are tested at scale.