Consent, Deepfakes and Why We Use Original Characters
AI image tools can produce a convincing picture of almost anyone. That is exactly why adult AI art needs a firm line: no real people. Not celebrities, not streamers, not cosplayers, not an ex, not a classmate, not a colleague. This guide explains the harm, the laws that now address it, and why original characters are the only sensible subject.
This is general information, not legal advice.
What we mean by “real people”
Any identifiable living person, including:
- Using a real person’s name in a prompt.
- Uploading someone’s photo to a face-swap, “undress” or “nudify” tool.
- Training or using a LoRA on someone’s face.
- Making a character “inspired by” a real person so closely that people would recognise them.
It does not matter whether the image is photorealistic or drawn in anime style. If it is recognisably a real person in a sexual context, they did not consent to it.
Why it is harmful
Non-consensual intimate images, real or fake, cause serious and well-documented harm:
- Humiliation and fear. Victims describe the experience as a violation, even when everyone knows the image is fake.
- Reputation and work. Images spread and resurface in searches, sometimes for years.
- Harassment and extortion. Fake images are used to threaten, blackmail or silence people, especially women, journalists and public figures.
- Minors. “Nudify” apps have been used on photos of school students, which is both abuse and a serious crime.
Saying “it’s obviously fake” does not undo any of this. The point of these images is usually the harm itself.
The legal landscape
Lawmakers have moved quickly since 2023. A few key examples:
European Union: AI Act
The EU’s Artificial Intelligence Act includes transparency obligations for deepfakes: people who deploy AI to generate or manipulate image, audio or video content that resembles real people, objects or places and would falsely appear authentic must disclose that it is AI-generated, and providers must mark synthetic output in a machine-readable way. These transparency rules apply from August 2026. The AI Act sits alongside existing EU and national laws on privacy, image rights and harassment, and the EU has also adopted rules requiring member states to criminalise non-consensual sharing of intimate images, including manipulated ones.
United Kingdom: Online Safety Act
The Online Safety Act 2023 added offences to the Sexual Offences Act 2003 for sharing, or threatening to share, intimate images of a person without consent, explicitly including images that have been made or altered to appear to show them. These came into force in January 2024. The Data (Use and Access) Act 2025 went further and made creating such sexually explicit deepfakes an offence, not only sharing them. The Act also places duties on platforms to tackle this content.
United States: TAKE IT DOWN Act
The TAKE IT DOWN Act, signed into federal law in May 2025, makes it a crime to knowingly publish non-consensual intimate imagery, including AI-generated “digital forgeries” of identifiable people, and to threaten to do so. It also requires covered platforms to remove such content within 48 hours of a valid request from the person depicted. Many US states have their own deepfake and image-based abuse laws, and some allow victims to sue.
Elsewhere
Australia, South Korea and a growing number of other countries have specific offences for sexual deepfakes. Even where no deepfake-specific law exists, existing laws on harassment, defamation, privacy and image rights often apply.
Why platforms block it
Responsible generators:
- Block prompts containing names of public figures.
- Do not offer “undress” or face-upload features for NSFW content.
- Remove models and LoRAs trained on real individuals.
- Respond to takedown requests quickly.
If a service sells exactly those features, it is selling a tool for abuse. Our generator guide lists what to check.
Copyright characters: a related but separate issue
Using named characters from existing anime and games is not the same harm as depicting a real person, but it brings its own problems: copyright and trademark issues, platform takedowns, and art that looks derivative. We use original characters and generic archetypes throughout this site for both reasons.
Why original characters are better anyway
Leaving real people out is not just about avoiding trouble. Original characters are better for the art:
- You own the design. Nobody can tell you your character is “wrong”.
- Consistency is easier. You control every detail, so you can build a character sheet and reuse it. See character design basics.
- The work is more interesting. An elf ranger with a backstory, scars and a signature weapon beats yet another lookalike.
- It is safe to share within the rules of the platform, because nobody is harmed.
If you are a victim
- Do not engage with the person posting or threatening.
- Document the content: URLs, usernames, dates, screenshots of the page (not downloading the image itself if you can avoid it).
- Report to the platform using its intimate-image or impersonation form.
- Use hash-matching services such as StopNCII.org, which lets adults create a digital fingerprint of an intimate image so participating platforms can block it, without uploading the image itself.
- Contact police or a support organisation in your country. In the UK, the Revenge Porn Helpline; in the US, the Cyber Civil Rights Initiative.
The short version
Real people: never. Original adult characters: always. It protects others, it protects you, and it makes for better art. For the related rules on age, read safety and age rules.