OpenAI has indefinitely put on hold its plan to allow erotic conversations in ChatGPT, the company confirmed on Thursday, whose initiative fueled concerns about its potential harmful effects and reputational risk for the world leader in generative AI. This decision, revealed by the “Financial Times”, comes at a time when the San Francisco AI giant is refocusing on the development of professional tools, in search of revenue.
In October, OpenAI boss Sam Altman defied critics by announcing the easing of ChatGPT restrictions in December, promising to “treat adult users like adults”. Five months later, OpenAI told the FT it wants to conduct long-term research into the effects of sexual conversations and emotional attachments before any decisions are made, acknowledging the current lack of “empirical evidence”.
“Lemon mode”
Contacted by AFP, OpenAI indicated that it had “nothing to add”. Some employees questioned the compatibility of the product, nicknamed “lemon mode,” with the company’s stated mission — that technology benefits all humanity. Investors also raised the high reputational risk given the expected commercial benefits, adds the British daily.
Technical obstacles also weighed on: the age verification system developed by OpenAI would have an error rate greater than 10%, according to a source cited by the FT, reviving fears of exposure of minors to explicit sexual content – a sensitive subject as the company faces several legal proceedings in the United States.
This abandonment is part of the refocusing announced Tuesday by Sam Altman to his teams, which led to the shutdown of Sora2, the social network for videos generated by AI. This reorganization aims to concentrate limited resources – computing power and capital – on the development of more lucrative professional tools and AI agents. Despite an increasing turnover – estimated at 20 billion $ (15.90 billion €) on an annual basis – OpenAI burns far more cash than it generates.
