Is It Illegal to Make a Deepfake?

The short answer is no, not automatically. Making a deepfake is not a crime in itself in most of the world. What determines whether a particular deepfake is legal is not the technology but four things: what it depicts, whether the person shown agreed to it, what it is used for, and where you and your audience are. A satirical video of a politician and a non-consensual intimate image of a private citizen are both deepfakes, and the law treats them very differently.

This guide explains where the lines fall as of 2026: what is generally legal to make, what is broadly illegal, how US federal and state law approach the question, how other major jurisdictions handle it, and the disclosure rules that increasingly apply even to lawful synthetic media. One note before starting: this is general information, not legal advice. Deepfake law is changing quickly and varies significantly by country and by US state, so for a specific situation you should consult a qualified lawyer.

The reason the answer is nuanced rather than a flat yes or no is that lawmakers have deliberately targeted harms rather than the tool. As one legal guide summarizes it, creating or distributing a deepfake is illegal in many circumstances but not all. The rest of this article is about which circumstances.

The Short Answer: Four Factors Decide Legality

Almost every deepfake law in force turns on some combination of the same four factors, and running a piece of synthetic media through them is the fastest way to gauge risk.

‍

‍

The first is content: what the deepfake actually shows. Sexual content and depictions of minors sit in the most heavily criminalized category, while a fictional landscape or a clearly invented character raises almost no issue. The second is consent: whether the real person depicted agreed to the use of their face, voice, or likeness. The third is purpose: whether the intent is to deceive, defraud, harass, or influence a vote, as opposed to entertain, comment, or create art. The fourth is jurisdiction: the rules differ sharply between countries, and within the US between states, and the location of both the maker and the audience can matter. A deepfake that is lawful in one place and for one purpose can be criminal in another.

‍

What Is Generally Legal to Make

A large share of synthetic media is lawful. Parody and satire generally retain protection, especially in the US under the First Amendment, on the logic that the audience is not being deceived because the work is recognizable as commentary rather than fact. Art, film, and fiction are similarly treated with a lighter touch, and legitimate commercial uses such as synthetic voiceovers, digital avatars, and virtual presenters are widely accepted when built on consent.

Consent is the single most powerful protection. A deepfake made with the documented, specific permission of the person depicted removes most of the legal exposure, which is why film studios and advertisers license likenesses before recreating them. The other recurring theme is disclosure: content that is clearly labeled as AI-generated is far less likely to run afoul of the law, because most regimes are aimed at deception, and a label defeats the deception. The important caveat is that these protections are not absolute. A parody exception is more likely to cover an obvious meme than a realistic clip designed to be believed, and even clearly artistic content increasingly carries a labeling obligation.

‍

What Is Generally Illegal

The prohibited categories are consistent across most of the world, even where the specific statutes differ. Non-consensual intimate imagery, meaning sexual or nude deepfakes of a real person made without consent, is now criminal across the US and much of Europe and Asia. AI-generated child sexual abuse material is illegal everywhere, with no artistic or satirical exception of any kind. Using a deepfake to commit fraud, to impersonate someone for financial gain, or to steal an identity is covered by long-standing fraud, forgery, and identity-theft laws regardless of whether a jurisdiction has a deepfake-specific statute. Defamation law applies when a fabricated video damages a real person's reputation. And deceptive political or election deepfakes are restricted in a growing number of places, though these laws collide most directly with free-speech protections.

‍

US Law: A Federal Baseline and a State Patchwork

For years the US had no federal deepfake law, leaving a patchwork of state statutes. That changed with the TAKE IT DOWN Act, signed in May 2025, which became the first federal statute directly targeting AI-generated intimate imagery. It makes it a federal crime to knowingly publish non-consensual intimate visual depictions of an identifiable person, and it expressly includes AI-generated content, which the statute calls digital forgeries. Penalties reach up to two years in prison when the victim is an adult and three years when a minor is depicted. Just as significantly, it imposes obligations on platforms: as of May 2026, any site or app hosting user-generated content must run a notice-and-takedown process, remove flagged material within 48 hours of a valid report, and make reasonable efforts to purge duplicates, with the Federal Trade Commission enforcing compliance. The law carves out exceptions for matters of public concern and for content the depicted person consented to.

Two further federal measures would extend protections if enacted. The DEFIANCE Act, which has advanced through the Senate, would give victims a federal civil route to sue creators and distributors, with statutory damages up to $150,000, or $250,000 where the deepfake is tied to sexual assault, stalking, or harassment. The NO FAKES Act would create a federal right over one's voice and likeness, with exceptions for satire, commentary, and news reporting. Both remain bills rather than law.

Below the federal level, the states have moved fastest and furthest. By mid-2026, somewhere around 45 to 48 states had enacted at least one deepfake statute, generally falling into three buckets: sexual imagery and child sexual abuse material, election communications, and voice or likeness rights. Tennessee's ELVIS Act protects a person's voice and likeness from unauthorized AI imitation. On elections, roughly 30 states restrict deceptive political deepfakes, most by requiring a disclosure label rather than an outright ban, and these are the laws most vulnerable to constitutional challenge: California's broad prohibition in AB 2839 was permanently blocked by a federal court in 2025 on First Amendment grounds, while its narrower disclosure-only law survived.

‍

Deepfake Law Around the World

Outside the US, the direction of travel is similar but the tools vary. The European Union takes a transparency-first approach: Article 50 of the EU AI Act, in force from August 2026, requires that deepfakes be labeled as artificially generated, with exceptions that lighten the obligation for evidently artistic or satirical work, and its penalties can reach tens of millions of euros. Alongside it, the Digital Services Act pushes large platforms to manage manipulated-media risks, and GDPR governs the use of a person's biometric data. Several member states criminalize non-consensual sexual deepfakes directly, including France and Germany.

The United Kingdom criminalized sharing intimate deepfakes under the Online Safety Act 2023 and then went further, making the creation of such images a standalone offense, with prosecutors clarifying in 2026 that the offense applies even to content made purely for personal use. China runs the strictest labeling regime, requiring both a visible label and embedded metadata on AI-generated content. South Korea has among the toughest sexual-deepfake laws in the world, where even possessing non-consensual sexual deepfakes can be criminal, a stance hardened after a large-scale abuse incident in 2024. Australia handles non-consensual intimate deepfakes through its eSafety regulator, and Denmark has proposed giving people a copyright-like control over their own face and voice. Across these jurisdictions, non-consensual deepfakes are now criminal offenses in more than 49 countries, a figure that has more than doubled since 2022.

‍

Disclosure: The Rule That Applies Even to Lawful Deepfakes

A crucial shift is that disclosure obligations increasingly apply regardless of whether the underlying deepfake is harmful. Under the EU AI Act, a deepfake must generally be labeled as AI-generated even when the content itself is perfectly lawful, and China already requires both visible and embedded markers. In the US, several states impose disclosure requirements on AI-generated political content, and provenance standards such as C2PA are becoming the common technical basis for these labels. The practical takeaway is that legality and disclosure are two separate questions. A consensual, non-deceptive, entirely lawful deepfake can still create legal exposure if it is published without the label a given jurisdiction requires.

‍

What This Means If You Make or Host Deepfakes

For an individual creator, the safe path follows directly from the four factors. Get documented consent from anyone real who is depicted. Avoid the criminal categories entirely, meaning nothing sexual involving a real person without consent and nothing involving minors, ever. Do not use synthetic media to deceive, defraud, or impersonate. Label your content as AI-generated, both because it is increasingly required and because it defeats the deception that most laws target. And check the rules in your own jurisdiction, since a state or national line can turn a harmless project into a violation.

For platforms and organizations, the obligations are now operational rather than theoretical. The TAKE IT DOWN Act's 48-hour takedown requirement, the EU's labeling rules, and the UK's removal duties all assume a platform can actually find the offending or unlabeled synthetic media at scale, which is a detection problem as much as a legal one. This is where automated deepfake detection supports compliance: DuckDuckGoose's DeepDetector analyzes images and video for the signatures of synthetic media, helping platforms identify deepfakes so they can meet takedown, labeling, and user-protection obligations rather than relying on manual review. Detection does not decide the legal question, but it is what lets an organization act on it in time.

Frequently Asked Questions

Is it illegal to make a deepfake?
Not by itself. Making a deepfake is generally legal, and legality depends on the content, whether the depicted person consented, the purpose, and the jurisdiction. Parody, art, and consensual or clearly labeled deepfakes are usually fine, while non-consensual intimate imagery, child sexual content, fraud, and some deceptive political deepfakes are illegal.

Are deepfakes for parody or satire legal?
Usually, especially in the US, where satire and parody receive strong First Amendment protection because the audience understands the work is commentary rather than fact. The protection is narrower for realistic content designed to be believed, and even satirical deepfakes may need an AI-generated label in some jurisdictions.

Is it illegal to make a deepfake of a celebrity or public figure?
It depends on the use. Commentary and satire are generally protected, but using a celebrity's likeness commercially without permission can violate right-of-publicity laws, and a sexual or defamatory deepfake is illegal regardless of the subject's fame. Several US states, including Tennessee, specifically protect voice and likeness from unauthorized AI use.

What does the TAKE IT DOWN Act make illegal?
It makes it a federal crime to knowingly publish non-consensual intimate images of a real person, expressly including AI-generated deepfakes. It also requires platforms that host user content to remove flagged material within 48 hours of a valid report and to make reasonable efforts to remove duplicates, with the FTC enforcing platform compliance.

Do I have to label a deepfake as AI-generated?
Increasingly, yes, depending on where you are. The EU AI Act requires most deepfakes to be labeled even when the content is lawful, China requires both visible and embedded markers, and several US states require disclosure on AI-generated political content. Disclosure and legality are separate questions, so a lawful deepfake can still break a labeling rule.

Is it illegal to make a deepfake for personal use?
It can be. Some laws, such as the UK's offense of creating intimate deepfakes, apply even when the content is never shared and is made purely for personal use. For most non-sexual, non-deceptive content involving no real person, personal use raises little issue, but sexual deepfakes of a real person are a serious exception.

Are deepfake laws the same in every country?
No. They vary widely. The US combines a federal law on intimate imagery with a patchwork of state statutes, the EU emphasizes labeling and transparency, the UK and South Korea criminalize sexual deepfakes strictly, and China imposes the most detailed labeling rules. Because the maker's and the audience's locations can both matter, cross-border content adds complexity.

What happens if you make an illegal deepfake?
Consequences range from civil liability and platform removal to criminal charges, depending on the content and jurisdiction. Non-consensual intimate deepfakes can bring prison time under federal and state law, fraud and defamation carry their own penalties, and platforms that fail to remove flagged content can face regulatory enforcement.

This article is for general information only and is not legal advice. Laws on deepfakes and synthetic media change frequently and differ by jurisdiction. Consult a qualified attorney about your specific situation.