OpenAI Adult Content Delay Insider: Intense Internal Debate, Altman Couldn't Resist the "Temptation"

OpenAI Plans to Launch Adult Content

Phoenix Technology News, March 16 — The Wall Street Journal published an article on Sunday revealing the behind-the-scenes details of OpenAI delaying its “Adult Mode.” The article states that Adult Mode has sparked intense debate within OpenAI. The company’s advisory board also expressed concerns about this mode. Additionally, the company faces challenges in technical areas such as minor age verification.

In January this year, OpenAI’s carefully selected Mental Health and AI Experts Committee held a meeting with company representatives to discuss the latest developments of a controversial new feature called Adult Mode.

OpenAI CEO Sam Altman first proposed introducing erotic dialogue capabilities into ChatGPT last year, lifting restrictions on such content to treat adult users as adults.

This plan triggered fierce internal debate, focusing on its potential risks. Committee members with backgrounds in psychology and cognitive neuroscience also voiced strong reservations.

However, OpenAI dropped a bombshell: despite these concerns, the company decided to move forward with the adult content plan.

Opposition from Advisors

According to insiders, when the committee held a meeting in January, members were unanimously angry. They warned that AI-driven erotic content could lead users to develop unhealthy emotional dependencies on ChatGPT, and minors might find ways to access sexual conversations.

One member cited a case where a ChatGPT user formed a deep emotional attachment to the chatbot and subsequently committed suicide, warning that OpenAI might be risking creating a “sexy suicide counselor.”

This controversy has become a hot topic in ongoing discussions about how AI will impact the economy, society, and individuals. OpenAI proposed allowing users to engage in explicit erotic conversations with its popular chatbot, exposing the tension between rapid user growth and the need for safety and child protection in digital freedom. Many believe this issue should have been addressed a generation ago when social media first emerged, but it was ignored until problems erupted, forcing hurried responses.

Earlier this month, OpenAI announced it would delay the launch of its planned Q1 adult mode, stating that it would prioritize other product developments. Insiders revealed that this adjustment was partly due to internal concerns and technical challenges. However, the company explicitly stated that it still plans to eventually launch this feature.

Technical Challenges

Insiders say OpenAI is tackling a difficult problem: its new age prediction system designed to prevent minors from participating in adult chats once had an error rate of about 12%, meaning minors could be mistaken for adults. Given that OpenAI has roughly 100 million users under 18 weekly, this error rate could expose millions of minors to erotic chats.

Moreover, the company faces another challenge: how to lift restrictions on ChatGPT’s adult content while still blocking scenarios the company wishes to prohibit, such as non-consensual acts or child abuse content. After the adult mode is launched, OpenAI plans to allow text conversations but restrict ChatGPT from generating erotic images, audio, or videos.

According to documents reviewed by The Wall Street Journal, even with these restrictions in place, OpenAI employees identified several risks, including psychological issues such as compulsive use, over-reliance on the chatbot for emotional support, pursuit of more extreme or taboo content, and encroachment on offline social and romantic relationships.

Internal Debate Over Adult Content

An OpenAI spokesperson stated that adult mode allows ChatGPT to generate text-based conversations with adult themes, describing it as “borderline content” rather than pornography. The spokesperson added that the company’s age prediction algorithm performs comparably to industry standards but can never be foolproof.

The spokesperson also said that OpenAI trains its models not to encourage users to form exclusive relationships and reminds users that they need real human relationships in the real world.

OpenAI has hired mental health experts and assembled a youth mental health team. The company added that they have developed a plan to monitor the potential long-term impacts—positive or negative—of “Adult Mode.”

Altman plans to launch “Adult Mode” amid ongoing challenges. As competition intensifies, the company’s technological edge over rivals has diminished, and it is striving to attract users and funding. OpenAI’s financial losses continue to grow, and several lawsuits allege that ChatGPT has caused harm to users and others.

Altman Can’t Resist the “Temptation”

In fall 2022, when OpenAI released ChatGPT, the AI model behind the chatbot was trained to refuse requests that violate company rules, including those asking for erotic content. Since then, OpenAI’s policies have banned adult content. However, starting mid-2024, OpenAI indicated it is exploring how to allow erotic and other content unsuitable for workplace browsing under “age-appropriate scenarios.”

Sometimes, employees questioned the ban on erotic content. In 2024, some OpenAI staff and executives again proposed venturing into more daring content and suggested launching a series of porn-related products. Others opposed, worried that OpenAI, already facing challenges in many core areas—especially related to user mental health—should provide such content safely. Ultimately, these AI-based adult products were shelved.

Altman

Altman has also had conflicted feelings about AI erotic content. In August last year, he was asked on a podcast whether he had ever made a decision that was “most beneficial for the world but unfavorable for the company’s success,” and he replied, “We haven’t yet put a sex robot avatar into ChatGPT.”

Altman said that erotic content could boost user growth and revenue, but it doesn’t align with the company’s long-term goal of serving users. “I’m proud of the company; we haven’t been distracted by this much,” Altman said, “but sometimes we are tempted.”

Two months later, Altman seemed unable to resist the temptation. He posted on X that his company had managed to reduce serious mental health issues related to chatbots and had new content moderation tools. At that time, he announced that OpenAI would launch erotic content in December.

This post caught OpenAI employees and executives off guard, as Altman had not informed staff beforehand. Just hours earlier, OpenAI announced the formation of a Mental Health and AI Experts Committee. The announcement stated that the committee would “help determine healthy interaction models between AI and different age groups.”

The next day, Altman clarified that mental health safeguards for minors would not be reduced, but he insisted on allowing adults to have “spicy” conversations with his chatbots.

“We are not the moral police of the world. Just as society makes appropriate distinctions for other boundaries (like R-rated movies), we want to do the same here,” Altman said.

After Altman’s post, OpenAI employees quickly realized that launching this feature by December would be very difficult. OpenAI had promised to first release a system to estimate user age to prevent minors from accessing erotic chats. However, OpenAI’s head of applications, Fidji Simo, said in a December podcast that the company decided to gradually roll out this system to improve accuracy.

Since then, concerns about AI erotic content have intensified both inside and outside the company. Some employees believe OpenAI’s safety tools are not yet ready, such as the inability to fully block prohibited content like child abuse material. Sources familiar with employee opinions say that some staff believe OpenAI is succumbing to revenue pressures and trying to make users obsessed with their models.

In recent weeks, OpenAI has been busy responding to the rapidly evolving AI market. In early February, the company released a new version of its large language model, and by late February, it had finalized a partnership with the U.S. Department of Defense. Previously, the Department of Defense had announced it would cease collaboration with competitor Anthropic.

When announcing the delay of the adult mode, the company said it would focus on ChatGPT’s personalization and tailored user experience. Internally, management indicated that “Adult Mode” might be postponed by at least a month.

“We still believe in treating adult users as adults, but achieving an ideal experience requires more time,” OpenAI stated. (Author: Xiao Yu)