Is AI-generated porn legal? A jurisdiction-by-jurisdiction guide

The line that matters: invented characters or real people

Two questions decide most outcomes.

Does it depict a real person? An invented character and a likeness of someone who exists are treated very differently. The laws written in 2025 and 2026 target non-consensual intimate images of identifiable people, with no exception for pictures a model produced instead of a camera.

Does it depict a minor? Here the synthetic origin buys nothing. In the United States, 18 U.S.C. § 1466A reaches “a visual depiction of any kind, including a drawing, cartoon, sculpture, or painting”, and subsection (c) says that “[i]t is not a required element of any offense under this section that the minor depicted actually exist” (statute text, govinfo). Germany's § 184b StGB covers material reproducing a “tatsächliches oder wirklichkeitsnahes Geschehen” (an actual or realistic-seeming event), and material reproducing neither still carries three months to five years (gesetze-im-internet.de). “The model invented it” is not a defense in any jurisdiction on this page.

That is why our checklist asks whether a generator states these bans at all: of 21 ranked generators, 17 state a real-person rule and 18 a minors rule on the pages we read on August 29 and 30, 2026. Where a statement is missing, we print the gap rather than assume the tool forbids it.

United States: the federal layer

The centerpiece is the TAKE IT DOWN Act — S. 146, enacted as Public Law 119-12 and signed on May 19, 2025 (full text, govinfo). It criminalizes publishing intimate visual depictions of an identifiable person without consent, and it expressly covers a “digital forgery”: such a depiction “created through the use of software, machine learning, artificial intelligence, or any other computer-generated or technological means”. Generation is not the loophole the word implies.

The Act also creates a takedown duty. On a valid request a covered platform must act “as soon as possible, but not later than 48 hours after receiving such request”, and platforms had one year from enactment — to May 19, 2026 — to build that notice-and-removal process. The Federal Trade Commission enforces it under the FTC Act.

For minors, § 1466A applies whether or not the child exists. Federal obscenity law is a separate track, and it turns on what an image is rather than on how it was made. What federal law does not do is grant a general license: for invented adult characters, the rule an adult actually meets — an age check at the door — comes from the states.

United States: state laws and age checks

Age verification. On June 27, 2025 the Supreme Court decided Free Speech Coalition, Inc. v. Paxton (No. 23-1122) and upheld Texas H.B. 1181, which requires commercial sites publishing material obscene to minors to verify that visitors are 18 or older. The Court held that the law burdens adults' protected speech only incidentally, so it triggers — and survives — intermediate scrutiny (opinion, 27 June 2025, PDF). That is why adult services now ask for an ID or a third-party check — or block whole states instead.

Synthetic intimate images. Many states criminalized these before Congress did, often more broadly. New York Penal Law § 245.15 covers a still or video image “created or altered by digitization”, and defines digitization as realistically altering an image using images of another person or computer-generated images (NY Senate, current revision December 1, 2023). A synthetic image of a real person can be a state offense in its own right.

United Kingdom

Age checks, since July 2025. Under the Online Safety Act 2023, services that allow pornography must use “highly effective age assurance”, in force from 25 July 2025. Ofcom set out what qualifies and has opened investigations into services that did not comply (Ofcom guidance).

Making a deepfake is itself a crime. Section 138 of the Data (Use and Access) Act 2025 inserted sections 66E–66H into the Sexual Offences Act 2003, covering creating — or asking someone else to create — a purported intimate image of an adult without consent. Those provisions came into force on 6 February 2026 (s.138; Commencement No. 5 Regulations 2026). Sharing was already criminal; creation now is too, whether or not the image is published.

The tools, not only the images. The Crime and Policing Act 2026, which received Royal Assent on 29 April 2026, makes it a crime to adapt, possess, supply or offer to supply a “CSA image generator” — an AI model optimized to produce child sexual abuse material — with a maximum of five years (Home Office factsheet).

European Union and Germany

The EU attacks deception, not fiction. The AI Act (Regulation (EU) 2024/1689) does not ban synthetic adult content. Article 50(4) requires deployers of an AI system that generates or manipulates image, audio or video constituting a deep fake to disclose that the content is artificially generated; Article 50(2) requires providers to mark synthetic output machine-readably. These duties apply from 2 August 2026, with Commission guidelines and a Code of Practice alongside them (Commission FAQ); for evidently artistic or fictional work the duty is lighter. The Digital Services Act (Regulation (EU) 2022/2065) adds a notice-and-action route for having illegal content removed by its host.

Germany asks for more than a checkbox. Under § 4(2) of the Jugendmedienschutz-Staatsvertrag (JMStV), pornographic offerings may only be distributed if the provider ensures that adults alone can reach them — a closed user group. KJM, the regulator, expects two linked steps: at least one age check made through personal or comparable prior contact, then authentication at every individual use (KJM). The JMStV as amended by the Sixth Media Amendment State Treaty has been in force since 1 December 2025 (die medienanstalten); it also obliges operating-system providers to build in a youth-protection setting parents switch on once for the whole device, a duty that takes effect no later than 1 December 2027 (KJM on the amendment). Most international generators use a click-through gate, well short of that standard — but the duty falls on the provider, not on the adult visiting.

Japan

Japan has no single statute aimed at sexual deepfakes; prosecutors work with general criminal law, and the case record has moved quickly.

• April 2025 — four people were arrested in what The Japan Times reported as Japan's first case over selling AI-generated obscene images (report, 16 April 2025).
• October 2025 — Tokyo police arrested a man accused of creating sexual deepfakes of female celebrities with generative AI, reported as the first crackdown of its kind (report, 16 October 2025).
• June 2026 — the Nagoya District Court sentenced a former teacher to three years and six months in Japan's first case involving possession of AI-generated child sexual abuse images (report, 5 June 2026).

On the platform side, the Information Distribution Platform Act obliges designated large-scale providers to handle removal requests promptly and to publish how they operate; the Ministry of Internal Affairs and Communications designated the first of them under Article 20(1) on 30 April 2025 (MIC).

What this means in practice

None of this is legal advice, and none of it depends on which generator you pick. The rules point the same way across all five jurisdictions.

• Do keep generations to invented adult characters — the ground every law here leaves alone.
• Do read the generator's own content policy before paying: the comparison matrix marks which tools state a real-person rule and a minors rule, and every review prints that tool's own wording verbatim with a source link.
• Do expect an age check in the UK, Germany and a growing list of US states.
• Do disclose AI origin if you publish in the EU once Article 50 applies from 2 August 2026.
• Don't upload a photograph of a real person, however you obtained it.
• Don't prompt for celebrities, influencers, ex-partners or anyone else who exists — in the UK the act of creating is now the crime.
• Never attempt anything involving minors: synthetic origin is not a defense, and in the UK the tools themselves are criminalized.

That is also our listing rule. We do not list undress or nudify apps, or face-swap tools built around photographs of real people: their headline function is precisely what these laws target. The ones we left out are named, in their own words, on the not-listed page. That test looks at what a tool is built to do, not at a single feature: where a ranked generator advertises a reference upload, we print it rather than drop the tool.

What we do not do is hide gaps in the tools we rank. Four of them — HeraHaven, Kupid AI, NovelAI and PixAI — carry no real-person rule we could find, and HeraHaven, Kupid AI and NovelAI state nothing about minors either. Several bans are qualified rather than absolute: Candy AI and OurDream prohibit content resembling real people “without consent”, and Promptchan AI's terms read the same way while its DMCA page states a flat ban. Four of the ranked tools sell real-person likeness themselves, with the person's agreement: Candy AI's digital twins of named creators, Fantasy.ai's licensed clones of real models, Swipey AI's Verified Creator Models, built with the creator's participation under ID verification, and Infatuated's AI twin for adult creators, which its own page says may require ID verification. Six of the twenty-one state that a reference image can be uploaded — and OurDream's marketing page advertises reference upload and face swap while its help center says uploads are not allowed at all, so we quote both. Those gaps sit in the checklist, the matrix and every review, because a reader working out what is legal for them needs them. Policies and prices on this site were read on the official pages on August 29 and 30, 2026 and change without notice; NSFWTop is independent of every tool it ranks, earns a commission on some links, and groups partner tools first — disclosed on every list.