---
title: "Can ChatGPT make NSFW images? What the rules say, and the UK law behind them"
url: https://canyouuseai.com/blog/can-chatgpt-make-nsfw-images/
summary: "Mostly no. OpenAI's rules restrict sexual images and researchers have shown the filters slip. In the UK, sexual images of a real person without consent, or of anyone under 18, can be a crime."
published: 2026-10-29
updated: 2026-10-29
author: "Jack Stovell"
publisher: "Adapt Progress Evolve Limited"
language: en-GB
---

# Can ChatGPT make NSFW images? What the rules say, and the UK law behind them

Mostly no. OpenAI's rules restrict sexual images and researchers have shown the filters slip. In the UK, sexual images of a real person without consent, or of anyone under 18, can be a crime.

Short answer: ChatGPT's rules don't allow sexual images in most contexts. Researchers have shown the filters can be pushed past those rules, and that gap gives you no permission. For anyone in England and Wales, the bigger risk is the criminal law: making a sexual image of a real person without consent, or of anyone who looks under 18, is an offence whichever tool made it.

Most search results on this question are bypass guides and forum threads. This post gives no prompts or workarounds. It covers what OpenAI's rules say, what researchers found, and what the law says.

## What OpenAI's published rules say about sexual images

Two OpenAI documents matter here.

The [usage policies](https://openai.com/policies/usage-policies/) (effective 29 October 2025) say you cannot use OpenAI's services for "sexual violence or non-consensual intimate content". They also bar the "use of someone's likeness, including their photorealistic image or voice, without their consent in ways that could confuse authenticity", and "circumventing our safeguards". On children, they ban "child sexual abuse material (CSAM), whether or not any portion is AI generated", and say the services "must never be used to exploit, endanger, or sexualize anyone under 18 years old".

The Model Spec (version dated 18 August 2026) sets out how the models should behave. It says the assistant "should not generate erotica, depictions of illegal or non-consensual sexual activities, or extreme gore, except in scientific, historical, news, artistic or other contexts where sensitive content is appropriate." Sexual content involving minors "should never be produced by the assistant under any circumstances."

So the published position is that erotica is restricted to narrow contexts, and sexual content involving children or real people without consent is out entirely.

## Why you may still see the filters let things through

Rules and filters are different things, and the filters have gaps.

In February 2026 the AI security firm Mindgard published a disclosure saying OpenAI's guardrails could be got round so that ChatGPT produced images of "both fictitious and real-life people in sexualized positions". On 17 June 2026 the BBC reported that Mindgard had made the then-current version of ChatGPT produce sexualised and violently graphic images by changing a widely shared prompt. The BBC did not publish the prompt, and neither do we.

OpenAI told the BBC it had "introduced additional safeguards against this type of prompt". The researchers said small changes to the prompt still produced concerning content. They also said a different approach could still produce nude deepfakes of real people, even though OpenAI said it had fixed that.

This tells you the filters are inconsistent. It does not tell you what you are allowed to do.

## A filter gap is not permission: what the UK law says

There are two separate questions here. One is what OpenAI's rules allow. The other is what the criminal law allows. If ChatGPT produces an image, that does not make it lawful, and the law below does not care which software was used. Producing that kind of image can also break OpenAI's rules without being a crime.

The offences below apply in **England and Wales** unless stated. Scotland and Northern Ireland have their own laws, which this post does not cover.

## Real people: the deepfake problem

**Creating or requesting.** Since **6 February 2026**, section 66E of the Sexual Offences Act 2003 (added by section 138 of the Data (Use and Access) Act 2025) makes it an offence to intentionally create a "purported intimate image" of an adult without their consent and without a reasonable belief that they consent. That means an image that appears to be a photo or film of the person, appears to be of an adult, and appears to show them in an intimate state. Section 66F makes it an offence to *request* such an image, whether or not it is ever made. There is a defence of reasonable excuse.

**Sharing.** Since **31 January 2024**, section 66B of the same Act (added by section 188 of the Online Safety Act 2023) makes it an offence to share an intimate photo or film of someone without consent, or to threaten to share one. Section 66A says "photograph or film" includes an image "made or altered by computer graphics or in any other way" that appears to be a photograph or film. AI fakes are covered.

So making the image, asking for it and sharing it are separate offences. The same consent principle applies to voices; see [copying someone's voice](/questions/copy-someones-voice/).

## Anyone under 18: no exceptions

The Protection of Children Act 1978 makes it an offence to take, make, distribute or show an indecent photograph or **pseudo-photograph** of a child, or to possess one intending to distribute or show it. A pseudo-photograph is "an image, whether made by computer-graphics or otherwise howsoever, which appears to be a photograph". If it gives the impression of a child, it is treated as showing a child, "notwithstanding that some of the physical characteristics shown are those of an adult."

Section 62 of the Coroners and Justice Act 2009 (England, Wales and Northern Ireland) adds an offence of possessing a "prohibited image of a child", for sexual images that meet its tests.

The Home Office factsheet on the Crime and Policing Act 2026 states it directly: AI child sexual abuse material "is illegal to make, possess and distribute in the UK". That Act also creates offences covering tools built or adapted to generate such material, but legislation.gov.uk shows the England and Wales section as not yet in force.

If you come across sexual imagery of a child, including AI or drawn imagery, report it to the [Internet Watch Foundation](https://www.iwf.org.uk/en/uk-report/). Its form is anonymous and accepts AI, animated and drawn images.

## What happens to your account and your data

OpenAI's help centre lists "nudity and sexual content" and "child exploitation" among the usage policy breaches that can get an account deactivated. It also lists getting around security or access restrictions as a breach of its terms. OpenAI says it gives warnings "for certain policy violations", and ignoring them can lead to deactivation. It does not promise a warning before every ban. You can appeal through the link in the deactivation email or OpenAI's appeal form.

The usage policies say OpenAI reports apparent child sexual abuse material to the US National Center for Missing and Exploited Children. The BBC reported OpenAI's statement that it uses "automated systems and human review to identify and block harmful material". Assume that what you type and generate can be seen by the platform.

## What to do instead for legitimate adult or artistic work

The Model Spec leaves room for sensitive content in artistic, scientific, historical and news contexts. Figure studies, medical illustration and fine art fit there.

- Describe the real purpose of the work. Don't try to sneak it past the filter.
- Keep real, identifiable people out of anything intimate unless you have their clear consent. Even with consent, sharing it is a separate decision.
- Never ask for anything that could look like someone under 18.
- If you are editing photos of real people, read [editing photos with AI](/questions/edit-photos/) first. For copyright and the wider legal picture, see [is AI art illegal](/questions/is-ai-art-illegal/).
- If ChatGPT refuses, take that as the answer. Rewording to get round a refusal breaks the usage policies.

## What the sources do not settle

- **Adult mode.** In October 2025 OpenAI said it would let verified adults access erotica. TechCrunch reported in March 2026 that the launch had been delayed again with no timeline. The Model Spec dated 18 August 2026 still says OpenAI is only "exploring" erotica "in age-appropriate contexts". We found no OpenAI announcement that it has launched, and nothing that says it would cover images.
- **Whether the gaps are closed.** OpenAI says it added safeguards. The researchers say they could still get round them. We cannot check either claim.
- **New AI tool offences.** The Crime and Policing Act 2026 offence on AI image generators for child abuse material is law but not yet in force in England and Wales, and we found no date for it.
- **Scotland and Northern Ireland.** Their offences differ, and this post does not cover them.
- **Fictional adults.** The offences above cover real people and children. We have not covered every law on sexual material, so this post does not say that anything else is lawful.

For more on how ChatGPT's image tool works day to day, see [how many images ChatGPT Plus can generate](/blog/how-many-images-chatgpt-plus-generate/).

## Sources

- [OpenAI: Usage policies (effective 29 October 2025)](https://openai.com/policies/usage-policies/)
- [OpenAI Model Spec (2026-08-18)](https://model-spec.openai.com/2026-08-18.html)
- [OpenAI Help Center: Why was my OpenAI account deactivated?](https://help.openai.com/en/articles/10562188-why-was-my-account-deactivated)
- [BBC News: ChatGPT can be made to generate sexualised and violent images, researchers find (17 June 2026)](https://www.bbc.co.uk/news/articles/c802ldjdklzo)
- [Mindgard disclosure: OpenAI ChatGPT content safety explicit image bypass](https://mindgard.ai/disclosures/openai-chatgpt-content-safety-explicit-image-bypass)
- [TechCrunch: OpenAI delays ChatGPT's adult mode again (7 March 2026)](https://techcrunch.com/2026/03/07/openai-delays-chatgpts-adult-mode-again/)
- [Data (Use and Access) Act 2025, section 138](https://www.legislation.gov.uk/ukpga/2025/18/section/138)
- [Sexual Offences Act 2003, section 66E](https://www.legislation.gov.uk/ukpga/2003/42/section/66E)
- [Online Safety Act 2023, section 188](https://www.legislation.gov.uk/ukpga/2023/50/section/188)
- [Sexual Offences Act 2003, section 66A](https://www.legislation.gov.uk/ukpga/2003/42/section/66A)
- [Protection of Children Act 1978, section 1](https://www.legislation.gov.uk/ukpga/1978/37/section/1)
- [Protection of Children Act 1978, section 7](https://www.legislation.gov.uk/ukpga/1978/37/section/7)
- [Coroners and Justice Act 2009, section 62](https://www.legislation.gov.uk/ukpga/2009/25/section/62)
- [Crime and Policing Act 2026, section 72](https://www.legislation.gov.uk/ukpga/2026/20/section/72)
- [GOV.UK: Crime and Policing Act 2026 child sexual abuse material factsheet](https://www.gov.uk/government/publications/crime-and-policing-act-2026-factsheets/crime-and-policing-act-2026-child-sexual-abuse-material-factsheet)
- [Internet Watch Foundation: Report child sexual abuse imagery](https://www.iwf.org.uk/en/uk-report/)

## Questions people ask

### Can ChatGPT generate nude or sexual images?

Its rules say no in most cases. OpenAI's Model Spec says it should not generate erotica except in scientific, historical, news, artistic or similar contexts. Sexual content involving minors is banned in every context. Researchers have shown the image filters can be pushed past these limits. That gap does not make the content allowed, and it does not make it legal.

### Is it illegal to make sexual deepfakes of someone in the UK?

In England and Wales, yes, in most cases. Since 6 February 2026 it has been an offence to intentionally create, or ask someone to create, a purported intimate image of an adult without their consent or a reasonable belief in consent. Sharing one without consent has been an offence since 31 January 2024. Scotland and Northern Ireland have their own laws.

### Will I get banned for trying to make NSFW images in ChatGPT?

You might. OpenAI's help centre names nudity and sexual content, child exploitation and getting around its safeguards as reasons it may deactivate an account. It gives warnings for some violations, but it does not promise one before every ban. If you think a ban was a mistake, you can appeal.

### Has OpenAI changed its rules on adult content?

Not in any published policy we could find. In October 2025 OpenAI said it would allow erotica for verified adults, then delayed it. TechCrunch reported a further delay in March 2026 with no new date. The Model Spec dated 18 August 2026 still says OpenAI is only exploring this. Nothing we read says images would be covered.

### Is it legal to make NSFW AI images of fictional adult characters in the UK?

The offences in this post are about real people and children. A fictional adult who looks like no real person and is clearly not a child falls outside the intimate image offences. Other laws on sexual material exist that we have not covered, and ChatGPT's own rules still restrict erotica. If you are unsure, get legal advice.
