OpenAI has reportedly pushed back plans for a more permissive “adult mode” for ChatGPT indefinitely, shelving the idea after concerns from employees and investors about what a sexualized or more erotically capable chatbot could do in the real world.
That it comes only days after OpenAI began closing down its Sora AI video maker makes it clear this isn't just another product pivot.
It looks a lot like a company realizing that the very core of AI's appeal is also what makes it impossible to be totally safe and free of controversy, especially at scale, no matter what CEO Sam Altman previously claimed.
Adulting AI
This was never just about whether ChatGPT should be allowed to get sexy or make videos. Both decisions reflect how adult-mode ChatGPT and Sora are in the uncomfortable category of AI that people are excited about right up until it starts making executives sweat.
Adult mode was always going to drag that contradiction into the light. The issue is not simply that sex is controversial - this was never just about sex. The issue is that the moment an AI system becomes convincing enough to sustain flirtation, sensuality, or even simply emotionally loaded companionship, it stops being easy to describe as a neutral tool. It becomes something more socially slippery, something much stranger and more fraught than a “helpful chatbot.”
A company can absolutely build a chatbot that is flirtier, more suggestive, and more open to adult themes. What it cannot realistically build is one that does all of that while also being guaranteed to remain harmless and uncontroversial in the mainstream future OpenAI is chasing.
Even without a formally announced IPO, OpenAI increasingly looks like a company thinking in those terms. And once a company starts imagining itself as a future public-market institution, its tolerance for weirdness tends to plummet.
A startup can flirt with cultural chaos and call it innovation. A company that may one day need to reassure investors every quarter starts looking at the same behavior and seeing risk exposure.
If OpenAI wants ChatGPT to become the AI layer for everything, it cannot afford for the product to be widely associated with endless titillation and scandal. It's mired in enough copyright lawsuits and fights with former founders as it is.
Maturity missing
The trap AI companies keep wandering into is that the features attracting users are the same ones that make them hard to domesticate.
Of course, plenty of adults are perfectly capable of engaging with erotic media or roleplay without falling apart or inviting public ridicule. The internet has contained stranger things for decades. And over-sanitized AI often feels absurdly juvenile, so why shouldn't mature adults have a mature AI?
But even if you accept all of that, it still does not solve OpenAI’s actual problem. The question was never whether adults could handle a more permissive chatbot. The question was whether OpenAI could. And the answer increasingly looks like no.
Not because every intimate or erotic use of AI is inherently catastrophic, but because the kind of product OpenAI wants to be in the world is fundamentally at odds with the kind of messiness a truly open-ended, emotionally convincing AI inevitably invites.
That is the bigger story here. ChatGPT can become more useful, more polished, and more integrated with our lives, but enhancing the chatbot enhances it's capacity for chaos, and no company eyeing Wall Street wants that.
Delaying adult mode is a confession that there is no perfect version of this product waiting just around the corner.
OpenAI can delay adult mode indefinitely. It can kill Sora. It can refocus on safer, broader utility. What it cannot do is solve the underlying contradiction with a settings toggle.
