This article is general information, not legal advice. Laws differ by jurisdiction and change fast; for decisions that matter, talk to a lawyer in yours.

Short answer: one principle organises almost everything: consent, and whether a real person is depicted. Fictional, AI-generated adults — characters who exist only as pixels — are legal adult content in most Western jurisdictions, subject to the same rules as any pornography (age-gating, distribution, obscenity edges). AI content depicting a real, identifiable person without consent is now criminal in the US federally and in 48 states, and unlawful across most of Europe. Around that line sits a growing layer of platform regulation: age verification with real fines in the UK, transparency and labeling duties in the EU. Anything involving minors, real or synthetic, is an absolute crime everywhere and an absolute block on any legitimate platform.

The United States: the Take It Down Act and 48 states

The federal picture changed on 19 May 2025, when the TAKE IT DOWN Act was signed into law. What it does:

• Makes it a federal crime to knowingly publish non-consensual intimate imagery (NCII) of a real person — explicitly including AI-generated “digital forgeries”. Penalties reach two years’ imprisonment for depictions of adults and three years where minors are depicted.

• Requires covered platforms to remove reported NCII within 48 hours of notice, giving victims a takedown mechanism with teeth.

Underneath the federal baseline, 48 states — every state except Ohio and New Mexico, as of August 2026 — have their own sexually-explicit-deepfake laws, many adding civil causes of action so victims can sue creators and distributors for damages. The trend in 2026 legislation is expansion of liability beyond individual creators to the platforms, payment processors and hosts that enable non-consensual content.

What this means in practice: generating or sharing sexual imagery of a real person without their consent — an ex, a celebrity, a streamer, anyone identifiable — is not a grey area, not “just AI”, and not low-risk. It is the most clearly criminalised act in this entire space.

The United Kingdom: age verification with enforcement

The UK’s Online Safety Act made age assurance mandatory for services showing pornographic content, enforceable since 25 July 2025 — and 2026 is the year enforcement became real:

• Ofcom has opened investigations into more than 90 services and issued multiple fines: £1 million against an adult site operator in December 2025, £800,000 against Kick Online Entertainment in February 2026, both for missing age checks.

• In January 2026 Ofcom opened investigations into X (over the Grok chatbot) and the AI companion service Joi.com — the clearest signal yet that generative-AI adult products are squarely inside the regime, not beside it.

• Accepted verification methods include facial age estimation, photo-ID matching, credit-card and open-banking checks and mobile-operator checks. Maximum penalties run to 10% of global turnover or £18 million.

For users, this is why UK visitors meet age gates; for anyone building in this space, it is why “we’ll add verification later” stopped being a viable plan.

The European Union: transparency and labeling

The EU AI Act’s Article 50 transparency obligations became enforceable on 2 August 2026. The parts that touch adult AI:

• Deepfakes must be labelled. AI-generated or AI-altered content that looks or sounds like a real person requires disclosure — intent to deceive is not required. Evidently artistic or fictional content needs only minimal, non-intrusive disclosure.

• Machine-readable marking. Providers must mark synthetic content so it can be detected; a Commission Code of Practice on marking and labeling (first draft December 2025) standardises the watermarking.

• Users must be told when they are interacting with an AI system — relevant to companion chat across the EU.

The practical upshot for fictional-character platforms is manageable (mark synthetic media, disclose the AI), and the upshot for real-person deepfakes is that they are now both illegal *and* required to self-identify — a combination designed to end the category.

The line that never moves

Every jurisdiction above, and every legitimate platform, treats any sexual depiction of minors — real, fictional, or “aged-up” — as an absolute crime and an absolute block. Synthetic does not launder it: US federal law, UK law and EU law all reach AI-generated material. There is no debate here and no platform worth using that treats it as one.

What this means for creators and users

• Fictional characters, clearly adult, clearly synthetic: the safe harbour. This is the category MyBabes operates in — generated characters, no real-person likenesses, hard blocks on the rest. The same logic explains policies you meet elsewhere: why Grok refuses real-person suggestive prompts, why we don’t clone voices without consent (Higgs Audio), and why “undress” tools aimed at photos of real people are legal poison regardless of what their landing pages say.

• Never generate sexual content of a real, identifiable person without documented consent. Not as a “risk”; as a crime with prison exposure in dozens of jurisdictions and a 48-hour takedown regime pointed at it.

• Expect age checks and treat platforms that verify as the compliant ones, not the annoying ones.

• Expect labeling to spread beyond the EU as the machine-readable marking standards mature.

• Jurisdiction matters. This article covers the US, UK and EU; parts of Asia, the Middle East and elsewhere prohibit pornography outright, synthetic or not.

Related in the Lab

• Grok Imagine and NSFW: how far it goes

• ElevenLabs and NSFW: the policy in practice

• ChatGPT and NSFW in 2026

Key takeaways

• Consent and real-person depiction are the organising line of every 2025–26 law.

• US: Take It Down Act (federal crime + 48-hour takedowns) plus deepfake laws in 48 states.

• UK: age verification enforced with real fines; AI chat and companion services now under investigation too.

• EU: deepfake labeling and machine-readable marking enforceable since 2 August 2026.

• Fictional adult characters remain legal adult content; minors are an absolute line everywhere.

FAQ

Is AI-generated porn legal?

Of fictional, adult characters: generally yes in the US, UK and EU, under normal adult-content rules. Of real people without consent: no — it is criminal in the US federally and in most states, and unlawful across Europe.

Is it illegal to make AI porn of a celebrity?

Yes, in the US under the Take It Down Act and state laws, and in most other Western jurisdictions. Publication triggers criminal liability and takedown duties.

Do AI porn sites have to verify age?

In the UK, yes, with substantial fines for failure; similar requirements are spreading elsewhere. Expect verification on any compliant platform.

Does the EU AI Act ban deepfakes?

It doesn’t ban them; it requires them to be labelled and machine-readable-marked from August 2026 — while non-consensual intimate deepfakes are separately illegal.

Last reviewed: August 2026. General information, not legal advice.