It was promised for December, delayed twice, then shelved indefinitely. Meanwhile the age checks shipped anyway. Where things really stand.
Adult mode never launched. OpenAI paused it indefinitely on March 26, 2026, and there has been no launch and no new timeline since. If you came here looking for the setting, there isn't one to find.
That is worth stating plainly, because a lot of pages ranking for this question say otherwise. Several currently describe adult mode as "in phased rollout" or offer steps to switch it on. Those steps describe a feature that does not exist.
The confusing part is that something did ship, and it goes the other way. OpenAI rolled out age prediction globally, which guesses whether you are under 18 and quietly applies teen restrictions if it thinks you might be. So a good number of adults arrived at this search after ChatGPT got more restrictive for them, not less.
| What you heard | What is true in September 2026 |
|---|---|
| Adult mode arriving December 2025 | Missed. Pushed to Q1 2026 |
| Adult mode arriving Q1 2026 | Missed again. Delayed March 7 |
| Erotica for verified adults | Shelved indefinitely on March 26 |
| Age verification | Live. Selfie and/or government ID via Persona |
| Age prediction | Live and global. Guesses your age from usage |
We build Secret Desires, which competes with ChatGPT for part of this, so treat the last section as an interested opinion. Everything before it is sourced and dated.
October 2025. Sam Altman announces the plan: as age-gating rolls out, and under a principle of treating adult users like adults, ChatGPT would allow more, including erotica for verified adults. The target was December.
December 2025. The window passes with nothing shipped. Altman reportedly sends an internal "code red" memo redirecting teams to the core ChatGPT experience.
January 2026. Age prediction begins rolling out globally, described by OpenAI as a way to determine when an account likely belongs to someone under 18. It is the age-gating half of the promise, arriving without the half users wanted.
March 7, 2026. TechCrunch reports a second delay. An OpenAI spokesperson says the company is pushing the launch out to work on higher priorities, adding that it still believes in treating adults like adults but that getting the experience right will take more time.
March 26, 2026. The Financial Times reports the project is paused indefinitely. TechCrunch's write-up notes the backlash came from watchdog groups and from OpenAI's own staff. The same week, OpenAI shut down Sora.
Since then. No launch, no revised date. OpenAI has said it wants more research into the psychological and societal effects of explicit AI conversation first.
Not squeamishness — the objections came from inside. Reporting points to three pressures arriving at once.
The Grok deepfake scandal ran through the same months, which did not help. Whatever you think of the decision, it was not made casually.
This is the part that actually affects you today. ChatGPT now estimates whether your account belongs to someone under 18, using signals like the topics you discuss and when you use it. If it decides you might be, it applies the teen experience automatically and reduces sensitive content.
It errs toward caution, which means it errs toward you. OpenAI has said that where there is doubt it defaults to the under-18 experience. Adults have reported being dropped into teen mode — including long-time paying subscribers — with newer accounts and late-night usage among the patterns that seem to trigger it.
If that happened to you, verification is the documented fix. Sign in on the web, go to Settings, then Account, then Verify age. It runs through Persona, a third-party provider, which may ask for a live selfie, a government ID, or both depending on your country. In Italy there is a 60-day window before some features are disabled.
Worth being clear-eyed: verifying removes teen restrictions and returns you to the normal adult experience. It does not unlock anything extra, because there is nothing extra to unlock. The trade is a selfie or an ID against getting back to baseline.
Roughly what it could before any of this. Frank discussion of sexuality, relationships, and medical topics in a non-explicit register. Romance and attraction in fiction. Fade-to-black.
What it won't do is sustained explicit content, and that has not changed. The jailbreak prompts circulating for this are worse than useless — modern refusals are semantic rather than keyword-based, so the old tricks break immediately and violate terms you agreed to. We don't publish them.
No. Paused indefinitely since March 26, 2026, with no announced timeline.
You can't. There is no setting, because the feature was never released. Any guide with steps for it is describing something that doesn't exist.
Age prediction estimates your age from account signals and defaults to the under-18 experience when uncertain. Verify your age in Settings, then Account to remove it.
Only if you want teen protections removed after being misclassified. You can keep using ChatGPT with them in place. Verification goes through Persona rather than to OpenAI directly.
Unknown. OpenAI has not ruled it out and has said it wants more research first. Given two missed dates and an indefinite pause, treat any specific date you see as speculation.
Not explicitly. Romance and non-graphic scenes are fine; sustained explicit writing is refused. We cover the fiction question on its own in our guide to Grok's NSFW policy, which is the closest any major lab currently gets to yes.
We build Secret Desires, so weigh this accordingly. The honest split.
Keep waiting on ChatGPT if you want one assistant for everything and adult content is a nice-to-have. It is the better general tool by a distance, and no purpose-built adult platform will match it for code, research, or analysis. If the feature ever lands, it will land there.
Stop waiting if the adult conversation was the point. What this timeline shows is that adult capability inside a general assistant is always the first thing cut when priorities shift — promised in October, dropped by March, with age checks shipping in between. That is not bad faith, it is what happens when explicit content is a side feature of a product with a billion other obligations.
Adult AI chat built for the purpose doesn't carry that risk, because there is no roadmap on which it gets deprioritised. NSFW AI that was designed for this from the start needs no verified-adult toggle that may never arrive, and for longer scenes, uncensored AI roleplay keeps characters consistent across sessions in a way a general chatbot does not attempt.
Where we are honestly worse: Secret Desires will not debug your code, read your PDFs, or summarise the news. It does one thing. If you wanted a single tool for everything, ChatGPT wins outright and this was never a real comparison. The memory problem that decides whether a long scene holds together is covered in how uncensored AI roleplay works.
Grok NSFW in 2026: adult text chat still works behind an 18+ setting, while image generation got locked down. The dated timeline and what stays blocked.
Read more
