OpenAI shelves erotic chatbot plans indefinitely amid safety fears, investor pushback

OpenAI has indefinitely shelved plans to launch an erotic chatbot, as the company pulls back from controversial product experiments and refocuses on its core artificial intelligence offerings, according to a report by the Financial Times.

The Sam Altman-led company had already delayed the rollout of its so-called “adult mode” amid internal debates over whether the feature should be scrapped entirely, the report said, citing people familiar with the matter.

The decision reportedly comes after mounting concerns from employees and investors about the broader societal risks of sexualised AI, including the potential to foster unhealthy emotional dependence and expose minors to explicit content.

An OpenAI spokesperson confirmed to the Financial Times that the erotic model is currently “on hold” with no clear timeline for release. The company said it plans to conduct long-term research into the impact of sexually explicit interactions and emotional attachments before making any product decisions.

Ethical concerns, internal pushback

The project had sparked unease within OpenAI, particularly around whether such a product aligns with its stated mission of ensuring AI benefits humanity.

“AI shouldn’t replace your friends or your family; you should have human connections,” a former senior employee told the Financial Times, adding that the issue contributed to their departure from the company.

Investors have also raised red flags, with some questioning the limited business upside of an adult-focused product compared to its potential reputational and regulatory risks.

The controversy comes at a time when global scrutiny of tech platforms’ impact on children is intensifying. Companies like Meta and Elon Musk-owned Grok are already facing legal challenges over alleged harms to minors, adding pressure on AI firms to tread carefully.

Technical hurdles add to delays

Beyond ethical concerns, OpenAI has also struggled with the technical complexities of building such a system.

According to the report, training AI models that are typically designed to avoid explicit content to instead generate nuanced adult conversations has proved difficult. Dataset challenges, including filtering illegal or harmful material such as incest or bestiality, have further complicated development.

The proposed feature, internally referred to as “Citron mode”, would likely require users to verify that they are over 18. However, OpenAI’s age verification tools have raised their own concerns.

The company recently introduced an age prediction system, but the report noted it has an error rate of over 10%, increasing the risk of underage access. OpenAI said the system performs in line with industry standards and that it is continuing to improve accuracy.

Shift away from ‘side quests’

The move to shelve the erotic chatbot reflects a broader strategic shift at OpenAI. The company has increasingly prioritised productivity-focused tools, including coding assistants and its flagship chatbot, as it works towards building an integrated AI “super app”.

As part of this shift, OpenAI is also winding down projects it considers non-core, includes Sora video generation model and social app.

The move is part of a broader strategic reorientation at OpenAI as it prepares for a potential initial public offering as early as the fourth quarter of this year.