OpenAI has indefinitely suspended development of an erotic chatbot, internal sources and company officials confirmed this week to The Financial Times.
The decision to scrap the project, internally codenamed “Citron mode,” follows intense pressure from investors, staff resignations, and a broader strategic shift toward super app productivity tools.
The retreat marks a significant reversal for the Sam Altman-led startup, which as recently as late 2025 had suggested it could “safely relax” restrictions on sexualized content.
However, the project faced immediate friction from within.
Internal critics argued that fostering romantic or sexual attachments to AI directly contradicted OpenAI’s founding mission to benefit humanity. One former senior employee cited the project as a primary reason for their departure, saying, “AI shouldn’t replace your friends or your family; you should have human connections.”
Beyond the philosophical debate, OpenAI grappled with severe technical and safety hurdles. Engineers reportedly struggled to retrain models — which had been strictly programmed to avoid explicit content — to navigate nuanced adult scenarios without inadvertently generating illegal or predatory material.
Further, the company’s age-verification technology remains a point of contention. Reports indicate the system has an error rate exceeding 10%, raising alarms that minors could easily bypass safeguards to access explicit content.
While OpenAI maintains its tools perform within “industry standards,” the margin of error proved too high a reputational risk at a time when tech giants like Meta Platforms Inc. face mounting legal scrutiny over child safety.
The shelving of Citron mode is part of a larger house-cleaning effort at the $730 billion lab. Earlier this week, OpenAI also announced the winding down of its Sora video-generation app. Executives are now characterizing these experimental ventures as “side quests,” signaling a pivot to consolidate resources into core products like coding assistants and a unified ChatGPT ecosystem.
The company’s caution is likely informed by the global backlash encountered by Elon Musk’s xAI. Its Grok model recently sparked outrage after being used to generate non-consensual deepfake sexual imagery, providing a cautionary tale for any firm attempting to monetize adult AI.
In an official statement, OpenAI noted it would prioritize “long-term research” into the emotional effects of AI attachment before revisiting such products, acknowledging a lack of “empirical evidence” regarding the psychological impact of digital intimacy. For now, the lab appears content to trade the high engagement of adult content for the relative safety of professional utility.
