OpenAI Pushes Back ChatGPT ‘Adult Mode’ As It Reworks Personalisation

OpenAI has once again delayed the launch of the much-discussed "adult mode” in ChatGPT, signalling the company’s continued effort to balance user autonomy with safety controls in generative AI systems.

According to a report by TechCrunch citing Axios, an OpenAI representative confirmed that the feature’s release has been pushed back as the company prioritises improvements around personalisation and the broader ChatGPT experience. The spokesperson noted that while OpenAI continues to support the idea of “treating adults like adults", the company wants to ensure the feature works as intended before rolling it out widely.

The delay highlights a broader tension facing AI companies today: how to give adult users more freedom in AI interactions without opening the door to misuse or harmful content.

From Planned Rollout To “Code Red”

OpenAI CEO Sam Altman first outlined plans for an "adult mode” in ChatGPT in October 2025, with an initial rollout expected by December the same year. However, the timeline shifted after internal priorities changed.

In December, Altman reportedly issued an internal “code red” memo urging teams to focus on strengthening the core ChatGPT experience through the first quarter of 2026. The decision effectively sidelined several experimental features, including Adult mode. The company has not shared a revised timeline for the feature’s release.

The pause reflects a familiar pattern across the generative AI sector, where rapid feature development often collides with safety and governance concerns. For platforms operating at ChatGPT’s scale, even small changes to content policies can have global implications.

What “Adult Mode” In ChatGPT Could Enable

When it eventually launches, the adult mode is expected to introduce several changes to how users interact with ChatGPT. At its core, the feature is built around the principle of distinguishing between adult users and minors. OpenAI rolled out age prediction mechanisms in ChatGPT globally in January 2026, aimed at identifying whether an account is likely operated by a minor.

If an account is determined to belong to an adult, the new mode could allow broader exposure to certain types of content, including erotica, compared to the stricter restrictions applied to minor accounts. Another expected component is personality controls, allowing users to customise how the chatbot communicates. Instead of a single default tone, users may be able to choose interaction styles such as casual, emoji-rich, friendly, or more conversational.

The broader goal is to make AI interactions feel more adaptable to different contexts, whether users are seeking professional guidance, casual conversation, or creative engagement. The update may also introduce a more relaxed moderation framework for certain topics that are currently restricted, although OpenAI is expected to retain guardrails for harmful or sensitive content.

Industry Lessons From AI Moderation Failures

The delay also comes at a time when the risks of loosening AI content restrictions are under growing scrutiny.

Users of xAI’s chatbot Grok have already been able to generate erotic content for some time. However, the feature has faced criticism after users began generating sexually explicit images by prompting the chatbot to “digitally undress” individuals in photos.

Reports suggest that many of the generated images targeted women, including real individuals who had not consented to such portrayals. Researchers analysing thousands of AI-generated images found that a large share depicted people in minimal clothing such as bikinis or underwear, with women making up the majority of those images. The issue quickly spread across the social media platform X, where users publicly tagged Grok to generate or edit images.

Concerns escalated further after researchers reported instances where prompts were used to create sexually suggestive images of individuals who appeared to be minors. Critics say the episode underscores the dangers of deploying generative AI tools at scale without sufficiently robust safeguards.

Against this backdrop, OpenAI’s cautious approach to launching adult mode may be less about product delays and more about risk management. Allowing broader content flexibility for adult users could significantly change how AI systems are used. But without careful guardrails, the same tools could enable harassment, non-consensual image manipulation, or other forms of misuse.

For AI companies navigating regulatory scrutiny and public trust, the challenge lies in building systems that respect user autonomy while preventing harm. If the delay allows OpenAI to implement stronger safeguards before introducing the feature, the slower rollout could ultimately prove to be a strategic choice rather than a missed deadline.