OpenAI Scraps Plans for Adult AI Chatbot Amid Internal and Investor Concerns OpenAI has confirmed it is halting development of an erotic chatbot for adults, a project that had been under consideration since last year. The decision follows reported unease among company employees and investors, marking the second product the AI firm has shelved this week after announcing the shutdown of its Sora video generation tool. The planned feature, internally referred to as Citron mode, was initially announced in October 2025 with a target release for December of that year. Its launch was already delayed as the company internally debated the merits and risks of releasing such a product. It is now on indefinite hold with no planned release date. According to sources familiar with the matter, technical challenges contributed to the decision. Engineers reportedly struggled to retrain AI models that were originally designed to avoid erotic content. A significant hurdle was ensuring the chatbot could filter out requests for illegal behaviors, such as bestiality or incest, which the models had not been previously trained to handle. In a statement, OpenAI cited a desire for more research, saying it wanted to study the long-term effects of erotic chats and user attachment to AI, noting a current lack of empirical evidence on the subject. The company also indicated a strategic shift to focus on its core productivity tools, like coding assistants, and move away from what it termed side quests like Sora and the adult chatbot. The concept for adult-oriented features emerged after OpenAI announced plans to implement stronger parental controls and automatic age detection for ChatGPT. At the time, CEO Sam Altman stated the company had always been cautious about unhealthy AI attachments but felt it could safely relax content restrictions in most cases. However, the plan faced significant internal and external pushback. Investors grew concerned, particularly in the wake of controversies involving rival AI models, such as xAI’s Grok, which faced backlash for generating deepfake nude images of real people and children. Staff members also expressed worries, with at least one senior employee reportedly leaving the company over the issue. That individual emphasized the importance of human connections, arguing that AI should not replace friends or family. A further complicating factor is the performance of OpenAI’s age-verification technology. This system was introduced following lawsuits from families who claimed ChatGPT harmed their children. Reports indicate the current technology has an error rate above ten percent, which could still allow a substantial number of underage users to access adult content. OpenAI has stated that this error rate is within industry standards and that work is ongoing to improve the system’s accuracy.
