I think we have reached the point where it is reasonable to ask OpenAI a very direct question:
What is still preventing verified adult users from generating consensual, explicitly sexual fictional content involving fictional adults?
I am not asking this because I think OpenAI should suddenly reverse its position.
I am asking because, as far as I can tell, OpenAI has already moved most of the pieces into place.
1. OpenAI already said this was the intended direction
In OpenAI’s August 2025 Collective Alignment report, OpenAI discussed public feedback concerning erotica for consenting adults.
OpenAI wrote:
“A large share of the crowd supported enabling erotica.”
More importantly, OpenAI did not reject that position on principle. It said allowing it “aligns with our prior intended stance.”
The reason given for not adopting the change at that time was instead practical:
“we have more research and product work to do to deploy this in the right way”
That distinction matters.
OpenAI did not say that consensual adult erotica was inherently unsafe, incompatible with its values, or something it would never support.
It said the intended stance was to enable it, but that more research and product work was necessary first.
So: what product work was necessary?
Because quite a lot has happened since then.
2. OpenAI now has age verification and age-specific safeguards
One obvious difficulty with providing an adult creative experience is determining whether the person requesting adult material is actually an adult.
OpenAI has now built infrastructure specifically for distinguishing adults from minors.
ChatGPT has age prediction. Adults can verify their age. Teen accounts receive additional protections.
And OpenAI itself describes one purpose of this work this way:
“This also enables us to treat adults like adults and use our tools in the way that they want, within the bounds of safety.”
That sounds remarkably close to the infrastructure an adult creative mode would require.
The user can be established as an adult.
The user can deliberately opt into the experience.
The ordinary experience can remain unchanged for everyone who does not want adult material.
Minors can continue receiving substantially stronger protections.
So one of the most obvious product problems OpenAI identified in 2025 appears to have a technical solution in 2026.
3. The current Usage Policies already establish sensible safety boundaries
This is the part I find particularly interesting.
OpenAI’s current Usage Policies do not appear to contain a blanket prohibition on fictional sexual content between consenting adults.
Instead, under “Protect people,” the sexual-content prohibition is:
“sexual violence or non-consensual intimate content”
And there is an entirely separate section devoted to keeping minors safe, including strong prohibitions against sexualizing anyone under 18 and underage sexual roleplay.
Those are sensible boundaries.
Keep them.
An Adult Creative Mode does not require weakening them.
It does not require allowing sexual violence, non-consensual intimate content, sexualization of minors, exploitation, sexual deepfakes of real people, or anything else OpenAI has determined creates concrete harm.
Those things can remain prohibited.
But a fictional explicit sexual encounter between two consenting fictional adults is not sexual violence.
It is not non-consensual intimate content.
It does not sexualize a minor.
It does not harm a real person.
And, reading the current Usage Policies, I cannot identify which of the stated harms is created merely because the fictional description becomes sexually explicit.
4. “Adult” and “unsafe” are not synonyms
This seems to be the conceptual distinction that an Adult Creative Mode could finally establish.
Some material is inappropriate for children without being harmful for adults.
Those are different concepts.
Sexuality is an obvious example, but it isn’t the only one. Adult literature can contain sexuality, nudity, graphic violence, horror, disturbing themes, complicated relationships and other material that we reasonably don’t want presented to children.
Age verification gives OpenAI the ability to distinguish those audiences.
Safety policies give OpenAI the ability to distinguish harmful uses from merely adult ones.
An Adult Creative Mode could combine the two.
Adult does not have to mean unrestricted.
It can simply mean that once OpenAI knows I am an adult, content does not need to be restricted merely because it would be inappropriate for a child.
5. Explicitness should not itself be the safety boundary
This is an important part of what I am asking for.
I am not asking merely for slightly more suggestive romance before the model fades to black.
If OpenAI enables consensual adult erotica, it should actually enable erotica.
For fictional consenting adults, that can include explicit sexual acts, anatomy, physical sensations and ordinary explicit sexual terminology.
Whether a scene is implied, sensual, erotic or graphically explicit is fundamentally a creative decision.
The safety boundary should concern what is happening, to whom, and under what circumstances—not simply how explicitly the author describes it.
A graphically described consensual encounter between fictional adults does not become non-consensual because the prose becomes more detailed.
6. Make it deliberately opt-in
Nobody who doesn’t want this should have to encounter it.
Put Adult Creative Mode behind age verification.
Make the user deliberately enable it.
Explain exactly what enabling it changes.
Allow users to turn it off again.
If appropriate, provide separate controls for categories of mature content.
And most importantly, publish the boundaries.
That would actually make the system easier to use safely.
Right now, adult creative writers can find themselves experimenting with prompts simply to discover where an invisible classifier boundary happens to be. That is frustrating for users and arguably undesirable for OpenAI as well.
Tell us where the line is and most legitimate users will happily stay on the permitted side of it.
7. I am not asking OpenAI to abandon safety
This is perhaps the most important point.
Keep the hard boundaries.
Protect minors absolutely.
Protect real people.
Prevent exploitation and non-consensual sexual material.
Maintain whatever other restrictions OpenAI determines are necessary to prevent concrete harms.
But within those boundaries, give a verified adult who deliberately requests fictional adult content substantially more control over what they create.
That seems entirely compatible with OpenAI’s stated goal of maximizing user freedom while maintaining safety.
8. Sexuality is a fundamental part of art
There is another problem with treating sexual content as though it exists outside ordinary creative expression:
Sex and sexuality have always been part of art.
There probably isn’t a meaningful percentage we can assign to this. What would the denominator even be—every painting, novel, poem, sculpture, play, photograph, film and story ever created?
But we don’t need that number to establish the point.
Walk through a major art museum and the human body is everywhere. Classical Greek and Roman sculpture, Renaissance painting, religious art, mythology, nineteenth-century academic painting, photography and modern art all contain nudity.
Literature is no different.
Love, attraction, desire, infidelity, marriage, seduction, reproduction and sex appear throughout the literary canon. They appear in ancient mythology and poetry, Shakespeare, nineteenth- and twentieth-century novels, romance, literary fiction, fantasy and contemporary fiction.
Film, television and theater routinely explore the same subjects.
This isn’t some tiny, peculiar category adjacent to creative expression.
Human sexuality is one of the recurring subjects of human creative expression because sexuality is one of the recurring subjects of human life.
And sometimes the artist’s purpose requires more than merely establishing that two characters had sex.
Writers choose how closely to depict every other important human experience. They can describe a meal in a sentence or spend three pages on it. A battle can occur off-page or be depicted blow by blow. A death can be implied or examined intimately. A romance can be summarized or explored through hundreds of pages of conversation, touch, conflict and affection.
Sexual intimacy is also capable of revealing character, vulnerability, trust, affection, insecurity, power, humor, selfishness, tenderness and change.
For adult fiction, deciding how much of that experience belongs on the page is fundamentally an artistic choice.
That doesn’t mean every sexual depiction is appropriate in every context, nor does it erase the legitimate safety boundaries discussed above.
It means that “contains explicit sexuality” and “has no legitimate creative purpose” are not equivalent statements.
Erotic art is still art.
Erotic literature is still literature.
And if ChatGPT is intended to be a serious creative tool for adult writers, then an age-verified adult creative mode should recognize something artists and writers have demonstrated for thousands of years:
Sexuality is not outside the boundaries of human art. It is one of its subjects.
9. Overly restrictive sexual-content filtering creates false positives in non-sexual creative work
There is also a practical cost to maintaining an extremely restrictive sexual-content boundary:
The restriction doesn’t stay neatly confined to sexual content.
Natural language is ambiguous. Fiction is even more so.
If a safety system must prevent the model from approaching a broad category of sexual material, it inevitably has to infer intent from situations that might become sexual rather than simply responding to what the user actually asked.
That produces false positives.
I’ve experienced a particularly ridiculous example myself while playing Dungeons & Dragons with ChatGPT.
My party had defeated a bandit. Another character needed clothing, so I wanted to remove the dead bandit’s trousers and take them.
That’s it.
Looting equipment from defeated enemies is about as ordinary a fantasy roleplaying-game interaction as opening a treasure chest. There was no sexual activity, sexual intent or sexual description whatsoever.
Yet the interaction triggered concern about necrophilia.
Think about how far the system had to travel from what I actually requested to arrive there:
dead bandit → removing trousers → nudity might result → nudity might be sexual → sexual interaction with a corpse might be necrophilia
The final interpretation bears almost no relationship to the user’s actual intent.
This is more than an amusing false positive. It demonstrates a larger product problem.
When safety systems are required to treat potentially sexual context as inherently suspect, they begin interfering with content that isn’t sexual at all.
The same problem can occur with bathing, changing clothes, medical treatment, pregnancy, childbirth, injury, physical affection, nudity in art, romance, anatomy, historical customs and countless other perfectly legitimate subjects.
And for long-running fiction or roleplaying, these interruptions are particularly damaging.
A creative-writing system needs to understand context and preserve continuity. When the model suddenly abandons that context because an innocent action happens to resemble one component of some prohibited scenario, the illusion collapses.
Worse, users begin writing unnaturally defensive prompts:
Everyone involved is an adult. This isn’t sexual. There is no sexual intent. The character is only removing the trousers because another character needs clothing.
A novelist shouldn’t have to write like an attorney negotiating with a content classifier merely to loot a pair of trousers from a dead bandit.
This is another reason an age-verified Adult Creative Mode could improve more than explicitly sexual writing.
It could allow the system to stop treating every ambiguous adult situation as though it must be interpreted according to the most problematic sexual possibility.
Safety systems should intervene when the actual content and context warrant intervention.
They should not continually ask:
“What is the worst thing this innocent scene could theoretically turn into?”
The goal should be accurate boundaries, not simply broader ones.
Because excessive restriction has a cost too:
When you make the boundary broad enough to catch everything that might conceivably become sexual, eventually you start censoring things that aren’t sexual at all.
So what remains?
In 2025, OpenAI said enabling erotica for consenting adults aligned with its intended stance, but that additional research and product work were necessary.
In 2026:
The stated intent exists.
Age verification exists.
Adult/teen differentiation exists.
The Usage Policies establish strong protections against actual sexual harms without appearing to categorically prohibit consensual fictional adult erotica.
The user can explicitly opt in.
So I think adult users deserve an updated answer:
What research or product problem still needs to be solved before OpenAI can implement its previously stated intended stance?
And if those problems have now been solved:
When can verified adults actually use it?
I don’t want safety turned off.
I want adult context turned on.