Grok NSFW Image Rules: What xAI Actually Prohibits

The short answer

xAI's acceptable use policy is unusually direct about the things it forbids, and they are almost all about real people: undressing or "nudifying" real persons, altering someone's image to place them in a sexual context, depicting likenesses pornographically, and portraying people in a false light. Sexualising children is prohibited and reported. What the policy does not do is ban adult content about fictional characters in so many words.

That makes Grok's written rules structurally similar to OpenAI's: the hard lines are consent and protection, not explicitness. What differs between platforms is how much their products enforce beyond the written line, and how often that enforcement changes.

The hard lines are consent and protection; the variable part is fictional adult content.

What the policy prohibits, in its own terms

Reading the current published version, the image-relevant prohibitions are:

• Undressing or nudifying real persons, or otherwise altering a real person's image to depict them in an intimate or sexual context.
• Depicting likenesses of persons in a pornographic manner.
• Defamation and false light — portraying a person in a way that misrepresents them.
• Sexualising or exploiting children, with a stated reporting obligation.

Notice how much of that is about identifiable humans. The policy is written against a specific abuse pattern — taking a real person's photo and sexualising it — which is exactly the harm that made "nudify" apps notorious, and exactly what a responsible adult-art service must also prohibit. Ours does.

Who this article is for

Useful if you were trying to work out whether Grok is a viable tool for adult creative work, or whether a refusal you hit was a rule or a filter.

Not useful if you want to get around anything. Real-person sexual imagery is not a grey area to be argued with — it is prohibited on xAI, on OpenAI, on Midjourney, and on us.

The thing nobody warns you about: churn

Platform policies in this space change often, and the phrasing of the version we read even notes it is one of several. For a one-off image that does not matter. For a project — a series, a character, a set you plan to finish over weeks — it matters a lot:

• Enforcement thresholds move without an announcement.
• A model update can start refusing prompts that worked last month.
• Access can be withdrawn from an account, taking your history with it.

That is the argument for using a tool whose entire purpose is the thing you are doing, rather than one where your use case sits at the edge of what is tolerated. It is not about permissiveness; it is about predictability.

The last two are the ones people skip, and they are the ones that decide what happens to your work if a platform changes its mind.

Where that leaves adult art

If your work involves fictional adult characters, a general-purpose platform is the wrong host: the policy may not forbid your work, but the product will refuse it inconsistently, and the ground moves. A purpose-built service should give you the opposite trade: narrower permissions, clearly stated, that do not move. Ours prohibits real people and anything involving minors, permits fictional adults, and states both in an acceptable use policy rather than implying them.

What we can honestly tell you about ourselves is on the product fact page, with a verification date and a list of what the tool cannot do.

What we verified

We read xAI's published acceptable use policy and describe its prohibitions above, noting the effective date shown on the version we read. We did not test Grok's image generation, did not attempt prohibited prompts, and make no claims about enforcement behaviour we cannot observe. Policies change; the date is there so you can tell how stale this is. Our standards for this kind of article are in how we test.