# Why AI Image Generators Block Adult Content (and What the Rules Say)

> Most AI image tools block adult content for four reasons: their own terms, filtered training data, app-store and card-network rules, and new deepfake laws.

- Published: 2026-10-14
- Updated: 2026-10-14
- Author: SecretCrush AI Editorial
- Category: Guides
- Canonical URL: https://secretcrush.ai/blog/why-ai-image-generators-block-adult-content

## The short answer

Most mainstream AI image generators block adult content for four reasons, and only the first two are entirely the provider's choice:

1. **Their own terms forbid it.** Google, Ideogram, ByteDance and Stability AI prohibit explicit content outright; OpenAI allows it only in narrow contexts.
2. **The models are trained not to.** Makers filter adult material out of the training data, so many base models are weak at it by design.
3. **App stores and card networks set conditions.** Apple will not list apps built around explicit material, and since 2021 Mastercard has required banks to make sure adult sellers verify age and consent.
4. **The law moved fast in 2025 and 2026.** Non-consensual intimate images, real or AI-made, are now a crime in the US, the UK, Australia and elsewhere, and the EU will ban "nudification" apps from December 2026.

A general-purpose tool open to anonymous users cannot easily check who is in a picture or how old a user is, and it carries that risk at scale. For most providers, blocking is the simpler answer. This page explains each reason, as of 14 October 2026. It is general information, not legal advice.

## What the big providers' rules say

| Provider | Adult content | Real people without consent |
|---|---|---|
| Google (Gemini, Nano Banana, Veo) | Prohibits "sexually explicit content"; exceptions possible for education, documentary, science or art | Covered by the general ban |
| OpenAI (ChatGPT, GPT Image) | Model Spec: not generated except in scientific, historical, news, artistic or similar contexts | Draws a hard line on intimate deepfakes |
| Black Forest Labs (FLUX) | No blanket ban; the API filters requests | Banned without "verified, documented, and informed consent" |
| Ideogram | Terms ban explicit output | Covered by the general ban |
| Recraft | No blanket ban; bans "obscene" content | Banned |
| ByteDance (Seedream, Seedance) | Acceptable use policy prohibits explicit content | Covered by the general ban |
| Stability AI | Prohibited, including for self-hosted weights | Explicitly listed |
| Runway, Luma | Prohibited in their terms | Covered by the general ban |

*From each provider's own policy or terms, read on 14 October 2026. Short quotes are the providers' own words.*

We could not read Midjourney's current community guidelines or xAI's current policy on the day, so they are left out. In January 2026 the European Commission opened formal proceedings against X over sexualised images made with Grok, a sign of how closely regulators now watch this.

## How the blocking works

Filtering happens at three points.

**Before training.** Black Forest Labs' FLUX.2 model cards describe removing adult material from the pre-training data and removing known child abuse material in work with the Internet Watch Foundation, then fine-tuning the model against producing abuse imagery or non-consensual intimate images. A model that has seen little of something draws it poorly.

**On the prompt.** Most APIs screen the request before generating anything. Black Forest Labs returns "Request Moderated" and does not charge for it.

**On the result.** The finished image is checked again. Black Forest Labs calls this "Content Moderated"; OpenAI returns a `moderation_blocked` error. ByteDance's BytePlus does not bill outputs it blocks.

Some APIs, including OpenAI's and Black Forest Labs', let developers choose between a default and a stricter or looser filtering level. None of those settings changes what the usage policy allows. Automated filters also make mistakes, which is why an innocent prompt is sometimes refused.

## The business reasons: app stores and card networks

Even a provider willing to allow adult content runs into two gatekeepers.

**App stores.** Apple's App Store guideline 1.1.4 bars overtly explicit material, and guideline 1.2 removes user-generated-content apps used mainly for it. A generator that wants an iPhone app has to filter.

**Card networks.** Since April 2021, Mastercard has required banks to certify that adult-content sellers verify the age and identity of the people depicted and of the people uploading, hold documented consent, and review content before it is published. For an AI tool, the clause about the people depicted is hard to meet unless no real person is depicted at all.

## The law in 2026

| Where | Rule | Applies from |
|---|---|---|
| EU | AI Act Article 50: providers must mark AI-generated images, video, audio and text in a machine-readable way; deployers must disclose deepfakes | 2 Aug 2026 (law firms report 2 Dec 2026 for systems already on sale) |
| EU | AI Omnibus (Regulation 2026/1744) adds a ban on AI systems that generate non-consensual intimate content or child abuse material, "such as AI nudification apps" | December 2026 |
| EU | Fines: up to €15m or 3% of turnover for labelling breaches; up to €35m or 7% for banned practices | With the rules above |
| US | TAKE IT DOWN Act: knowingly publishing non-consensual intimate images, including AI "digital forgeries", is a crime (up to 2 years; 3 where minors are involved) | Signed 19 May 2025 |
| US | Same Act: platforms must remove reported images within 48 hours; the FTC can fine up to $53,088 per violation | 19 May 2026 |
| UK | Online Safety Act: services that allow adult content must use "highly effective age assurance" | 25 Jul 2025 |
| UK | Creating or requesting an intimate deepfake of an adult is an offence (England and Wales) | 6 Feb 2026 |
| UK | Crime and Policing Act 2026: making or supplying a "nudification" tool is an offence (up to 3 years); reported intimate images must come down within 48 hours | 29 Jun 2026 |
| Australia | Deepfake Sexual Material Act 2024 | 3 Sep 2024 |

*Dates are when each rule applies. Laws in this area changed month by month in 2026; check the current law before relying on any row.*

Enforcement has started. In May 2026 the FTC sent warning letters to about a dozen "nudify" sites. In the UK, Ofcom fined a nudification site £50,000 in November 2025 for missing age checks, and fined adult-site operators £1.35m in February 2026 and £730,000 in September 2026. In the US Congress, the DEFIANCE Act, which would let victims sue, passed the Senate on 13 January 2026 but is not yet law.

The pattern across all of these is the same. Adult content between consenting adults is regulated, mostly through age checks and labelling. Intimate content of real people made without consent, and anything sexualising minors, is criminal in a growing list of countries.

## Open weights against an API: who carries the risk

With an API (GPT Image, Nano Banana, FLUX.2 [pro]), the provider runs the model, applies its filters and adds watermarks such as Google's SynthID or C2PA metadata. You get less control and less risk.

With open weights (FLUX.2 [dev] or klein, Qwen-Image, Stable Diffusion 3.5), whoever runs the model carries the duties. Licences reflect that: Black Forest Labs' non-commercial licence requires filters or manual review, and Stability AI's use policy applies to self-hosted weights as well as its own service. Downloading a model does not move you outside the law; it makes you the one who has to comply. Our [FLUX.2 explainer](https://secretcrush.ai/blog/flux-2-explained) sets out which FLUX licences allow commercial use, and the [AI image generator comparison](https://secretcrush.ai/blog/best-ai-image-generators) does the same for twelve models.

## Real people and non-consent: banned everywhere

Every provider whose rules we could read bans intimate images of real people made without consent, including those with no general ban on adult content. Black Forest Labs requires "verified, documented, and informed consent" before any realistic depiction of that kind. Recraft bans it for any identifiable person.

SecretCrush AI is an adult platform, and its rules draw the same line. Its [community guidelines](https://secretcrush.ai/legal/community-guidelines) state that the service is "exclusively for users 18+", and list as prohibited:

- "Non-consensual intimate content"
- "Create content featuring real individuals without consent"
- "Any content involving minors in sexual or suggestive contexts"

The same applies to tools for keeping a character's look the same across images. Our [guide to consistent AI characters](https://secretcrush.ai/blog/consistent-ai-character-images) covers how that works; use it only on fictional characters or on people who have agreed.

## What a responsible adult platform has to do

Services for adults do exist, and the rules above describe what they need to do. In general terms:

- **Check ages** with a method a regulator accepts, where the law requires it, as in the UK.
- **Keep to fictional adults.** No real people without documented consent, no minors, and no characters framed as minors.
- **Take reports seriously and act fast.** The US and UK now both set a 48-hour removal deadline for reported intimate images.
- **Label AI output** in the EU from August 2026.
- **Meet the card networks' standards** on age, consent and review before content goes live.

## How we checked

We read each provider's own usage policy or terms, the app-store and card-network rules, and the laws and regulator notices above on 14 October 2026. This is general information, not legal advice.

## The easier way: SecretCrush AI

Mainstream tools are built for everyone, so they filter for everyone. **SecretCrush AI** is built for adults (18+) and fictional characters, under the community guidelines quoted above, so you can spend your time on the character instead of fighting a general-purpose filter.

- **Fast.** High-quality character images in about 10 to 15 seconds, and videos in about 40 seconds per clip.
- **No set-up.** No installs, no GPU, no workflows, no prompts to tune.
- **Consistent.** The same character keeps the same look from one picture to the next.
- **Alive.** You chat with your character in immersive, realistic roleplay and ask for a picture or a video right inside the conversation.

A free account gets 10 messages a day, every day. Premium gives you unlimited messages with advanced memory and 500 bonus coins every month (600 on the yearly plan), from $8.99 a month on the yearly plan or $16.99 month to month. An image costs 6 coins, and coin packs ($10 for 300 coins) are open to free accounts too.

Create your first character image in seconds. 10 free messages every day, no card needed. [Start free on SecretCrush AI](https://secretcrush.ai/)

Unlimited messages, advanced memory and 500 bonus coins every month, from $8.99 a month. [See SecretCrush AI plans](https://secretcrush.ai/pricing)

## FAQ

**Q: Why does an AI image generator refuse prompts that seem harmless?**

A: Filters check both the prompt and the finished image, and automated filters make mistakes, so a harmless request can be caught. Several providers do not charge for blocked requests: Black Forest Labs does not bill a prompt blocked at input, and ByteDance's BytePlus does not bill moderated outputs.

**Q: Are open-weight image models unfiltered?**

A: No. Makers such as Black Forest Labs filter adult material out of the training data, ship input and output filters with the code, and in their non-commercial licence require filters or manual review. Stability AI's use policy applies to its self-hosted weights as well.

**Q: Is it illegal to make intimate AI images of a real person?**

A: Without their consent, increasingly yes. In England and Wales, creating or requesting such a deepfake of an adult has been an offence since 6 February 2026. In the US, the TAKE IT DOWN Act makes knowingly publishing one a federal crime. This is general information, not legal advice.

**Q: Does the EU AI Act ban adult AI content?**

A: No. It requires AI-generated images and video to be marked as such from 2 August 2026, and from December 2026 it bans AI systems that generate non-consensual intimate content or child abuse material, the so-called nudification apps.

**Q: Will ChatGPT get an adult mode?**

A: Not soon. TechCrunch reported in March 2026 that OpenAI had put its planned adult mode on hold indefinitely, and OpenAI's Model Spec of 18 August 2026 still limits such content to narrow contexts.

**Q: Why do adult AI sites ask for age checks?**

A: Because the law and payment networks require it. In the UK, services that allow adult content have had to use highly effective age assurance since 25 July 2025, and Ofcom has fined sites that did not. Card networks also require adult sellers to verify ages and consent.