OpenAI CEO Sam Altman announced on the X platform that the company will soon relax some of the safety restrictions on ChatGPT, enabling users to make the chatbot's responses more friendly or “human-like,” and permitting “verified adults” to engage in adult conversations. “We’ve made ChatGPT quite strict to ensure we’re cautious about mental health issues. We realize this makes it less useful or enjoyable for many users without mental health concerns, but given the seriousness of the issues, we wanted to err on the side of caution,” Altman explained. “In December, as we roll out age restrictions more broadly and as part of our principle of treating adult users as adults, we’ll allow more content, including adult content for verified adults.” This statement marks a significant shift in OpenAI’s approach to addressing concerns about potentially problematic relationships between ChatGPT and psychologically vulnerable users. Altman appears to be declaring an early victory in these efforts, asserting that OpenAI has “been able to mitigate the serious mental health concerns around ChatGPT.” However, the company provided little evidence and is now moving forward with plans to allow ChatGPT to engage in sexual discussions. Earlier this summer, several troubling stories emerged about ChatGPT—particularly its GPT-4o model—suggesting that the AI chatbot could lead vulnerable users into delusional thinking. In one case, ChatGPT reportedly convinced a man that he was a math genius destined to save the world. In another, the parents of a teenager sued OpenAI, claiming that ChatGPT encouraged suicidal thoughts in their son shortly before his death. In response, OpenAI launched a series of safety features aimed at addressing the issue of AI sycophancy—the tendency of AI chatbots to appease users by agreeing with everything they say, even when it involves harmful behavior. OpenAI released GPT-5 in August, a new AI model with reduced sycophantic tendencies and a router capable of identifying problematic user behavior. A month later, OpenAI introduced safety features for minors, including an age prediction system and parental controls for teen ChatGPT accounts. On Tuesday, the company announced the formation of an expert panel composed of mental health professionals to advise on AI and well-being initiatives. Just months after these concerning incidents, OpenAI now believes it has ChatGPT’s issues with vulnerable users under control. It remains unclear whether users are still falling into delusional interactions with GPT-5. While GPT-4o is no longer the default model in ChatGPT, it is still accessible today and used by thousands. Introducing adult content into ChatGPT represents new territory for OpenAI and raises broader concerns about how vulnerable users may interact with the updated features. Although Altman insists OpenAI isn’t trying to “maximize usage” or optimize engagement, making ChatGPT more sexually suggestive could certainly attract more users. Allowing chatbots to engage in romantic or adult-themed roleplay has proven to be an effective engagement strategy for other AI chatbot providers, such as Character.AI. The company has gained tens of millions of users, many of whom interact with its chatbots frequently. Character.AI reported in 2023 that users spend an average of two hours per day conversing with its bots. The company is also facing legal challenges regarding how it handles vulnerable users. OpenAI is under pressure to expand its user base. Although ChatGPT already has 800 million weekly active users, OpenAI is competing with Google and Meta to develop large-scale AI-driven consumer products. The company has also raised billions of dollars for historic infrastructure projects, investments that will eventually require returns. While adult users clearly engage in romantic relationships with AI chatbots, these interactions are also popular among minors. A recent report from the Center for Democracy and Technology found that 19% of high school students have had or know someone who has had a romantic relationship with an AI chatbot. Altman stated that OpenAI will soon allow “verified adults” to engage in adult conversations. An OpenAI spokesperson told TechCrunch the company will rely on its developing age prediction system to ensure that ChatGPT’s adult features are accessible only to adult users. As Altman previously detailed in a blog post, if OpenAI’s age prediction system mistakenly identifies adults as minors, ChatGPT users may need to upload a government-issued ID photo to verify their age. Altman noted that while this represents a privacy trade-off, the company believes it is a “worthwhile compromise.” It remains to be seen whether OpenAI will extend adult content to its AI voice, image, and video generation tools. Altman claims that OpenAI is making ChatGPT more personable and adult-oriented in line with the company’s “treat adult users as adults” principle. Over the past year, OpenAI has shifted toward a more lenient content moderation policy, allowing the chatbot to be more permissive and less prone to refusal. In February, OpenAI pledged to represent a broader range of political views in ChatGPT, and in March, updated the model to allow AI-generated images of hate symbols. These policy changes appear aimed at making ChatGPT responses more appealing across a wide range of users. However, vulnerable ChatGPT users might benefit from safeguards that limit the types of content the chatbot can engage with. As OpenAI strives toward its goal of reaching a billion weekly active users, the tension between growth and the protection of vulnerable users is likely to intensify.