Quick answer: There is no general federal deepfake law in the US as of April 2026, though the TAKE IT DOWN Act covers non-consensual intimate imagery. States cover three more categories: election deepfakes (California's version was blocked on First Amendment grounds), voice and likeness rights, and financial fraud under existing wire fraud statutes. Synthetic people fall outside most statutes.
Last updated: April 30, 2026.
The short answer is: it depends on who is in the video, what it is used for, and which state you live in. There is no general federal deepfake statute in the United States, though one narrow federal law (the TAKE IT DOWN Act) now covers non-consensual intimate imagery. What exists beyond that is a patchwork of state statutes, most of them narrow, and a few of them already struck down in court.
This post breaks down what the law actually covers, where the gaps are, and why legal protection is slower than the problem it is supposed to solve.
The Law Covers a Few Narrow Categories, and Everything Else Falls Through
Before getting into statutes, it helps to understand what "deepfake" means legally. If you need the technical grounding first, the guide to what a deepfake actually is covers the mechanics.
Legally, "deepfake" is not a defined term in most statutes. Laws address narrower things: non-consensual intimate imagery, election interference, voice and likeness rights, or fraud. A deepfake that does not fit one of those categories often falls into no category at all.
That framing matters more than the specific statutes.
Non-Consensual Intimate Imagery Has the Most Coverage
If a deepfake sexualizes a real person without their consent, most states have laws that apply, and as of May 2025 there is a federal statute too. Non-consensual intimate imagery (NCII) statutes have been on the books in various forms for years. Several states have updated them specifically to cover AI-generated content. The federal TAKE IT DOWN Act, signed in May 2025, criminalizes creation and non-consensual sharing at the national level and forces platforms to remove reported content within 48 hours starting May 19, 2026. For the full breakdown of how victims can use the federal law, see the federal law that forces platforms to remove your deepfake in 48 hours.
Some state laws carry criminal penalties. Others create civil liability only. The coverage is real, but it is not uniform across states, and enforcement requires identifying the creator. When content is generated anonymously and uploaded through a VPN, that process can take months or fail entirely. The TAKE IT DOWN Act narrows that gap by giving victims a federal removal right that does not depend on identifying the creator.
The law exists. The enforcement pipeline often does not, but it is closing.
Election Deepfakes: The Laws Are Already Fraying
Several states passed laws targeting deepfake political ads ahead of the 2024 election cycle. California's attempt (AB 2655) was blocked in October 2024 in Kohls v. Bonta and struck down on the merits in 2025. The court found it violated the First Amendment.
That ruling is significant. It signals that broad prohibitions on political deepfakes face real constitutional risk. Courts are likely to apply heightened scrutiny to any law that restricts political speech, even synthetic political speech.
The Federal Election Commission has not issued binding rules on AI-generated political content as of April 2026. In the meantime, deepfake political ads are circulating in the 2026 midterm cycle with no consistent legal check on them.
Financial Fraud Is Covered, But Not Under Deepfake Law
Deepfakes used to commit financial fraud are already illegal under existing wire fraud statutes and FTC consumer protection rules. You do not need a dedicated deepfake law to prosecute someone who uses a synthetic video to defraud investors.
Celebrity deepfake crypto scams are the clearest example. The deepfake is the method. The fraud is the crime. Prosecutors use the fraud statute, not a deepfake-specific one.
This is one area where the existing legal framework is not actually broken. The problem is attribution, not coverage.
The State-by-State Coverage Map (April 2026)
Pulling the categories together, here is what state coverage looks like across the country as of April 2026:
| Category | States with statutes (representative examples) | Status as of April 2026 |
|---|---|---|
| Non-consensual intimate imagery (AI-generated) | California (AB 602), New York (Civil Rights Law § 52-c), Texas (Penal Code 21.165), Virginia (§ 18.2-386.2), Florida, Illinois, Minnesota, Tennessee, Georgia, Washington, plus 20+ others | Most active category. Federalized by the TAKE IT DOWN Act in May 2025; platform compliance becomes enforceable May 19, 2026. |
| Election deepfakes | California (AB 2655, blocked October 2024 in Kohls v. Bonta), Texas (SB 751), Minnesota, Michigan, Washington, Wisconsin, Oregon, New Mexico, plus ~15 others | Patchwork. California's AB 2655 was preliminarily enjoined in 2024 and struck down on the merits in 2025 on First Amendment grounds. Broad versions face constitutional risk. |
| Voice and likeness rights | Tennessee (ELVIS Act, effective July 2024), New York, Washington, Illinois | Newest cluster. Tennessee's ELVIS Act was the first dedicated voice-cloning law in the US. Other states are following. |
| Financial / wire fraud (deepfake as method) | All 50 states; federal wire fraud (18 USC § 1343) | Already covered. No deepfake-specific statute is needed for prosecution. |
| Fully synthetic personas (no real person) | None | Gap. No state has addressed this yet. |
The coverage is uneven, and the state you live in often matters more than what the deepfake actually does. A non-consensual intimate deepfake in California has years of statute behind it. The same deepfake in a state without an updated NCII statute gets prosecuted under harassment or stalking law instead, with weaker penalties and a higher proof bar.
For the federal pathway specifically, the TAKE IT DOWN Act primer walks through the four legal options available to victims: federal criminal, federal civil, state criminal, and state civil.
170 Laws enacted across US states and the EU since 2022, with 146 new bills introduced in 2025 alone. Source: Public Citizen deepfake legislation tracker
The EU Is Moving Faster Than the US
The EU AI Act has moved from regulation to enforcement. Its guidelines now require certified detection infrastructure for large platforms operating in the EU. Companies deploying AI-generated content at scale need to detect and label it, or face compliance exposure.
This is a structural difference from the US approach. The EU is regulating the platforms. The US is regulating specific harmful uses. Both have gaps, but the EU framework creates a broader obligation for the technology industry, not just for bad actors.
Three Things the Law Does Not Cover
This is where the legal framework breaks down most visibly.
Fully synthetic people. If a deepfake does not use a real person's likeness, there is no identity theft claim. A fraud operation built on entirely fictitious AI-generated faces falls outside most deepfake-specific statutes. Courts have not resolved how existing fraud law applies here.
Satire and parody. First Amendment protection for satire is well-established. A clearly labeled parody deepfake of a politician almost certainly falls within protected speech. Lawyers genuinely disagree on where that line sits in ambiguous cases.
AI-generated content with no real person involved. Synthetic actors, synthetic news anchors, synthetic influencers. None of that is covered by laws written to protect real people's likeness rights. The technology has already moved past the framing the laws were written for.
Think you found an AI video?
Paste the URL and let the Ledger community verify it. Free.
Even Where Laws Exist, Enforcement Is Hard
Honest admission: even lawyers who specialize in this area disagree on how specific statutes apply to specific cases. The technology is new. The case law is thin. Jurisdictional questions get complicated fast when a deepfake is created in one country, hosted in another, and viewed in a third.
Enforcement requires identifying a creator who may be anonymous, pursuing a platform that may be uncooperative, and moving through a legal process that takes months. By the time a case resolves, the content has already done its damage.
This is not an argument against the law. It is an argument for not treating legal protection as the first line of defense.
Detection Moves Faster Than Courts
The legal system is reactive. It acts after harm. Detection can act before a video spreads.
When a community flags an AI-generated account on Ledger, that record persists even if the platform removes the content. Platforms remove videos. They do not always remove accounts. And removed content still spreads through downloads and reposts before it disappears.
A community-built record of flagged accounts builds pressure and documentation faster than a legal complaint. That does not replace the law. It fills the gap while the law catches up.
If you want to see how detection works in practice, the guide to spotting AI-generated TikTok videos covers the signals worth looking for before you report anything.
Related Posts
- What Is a Deepfake? A Plain-English Guide for Social Media Users: the technical and definitional grounding for how synthetic media is created
- AI Deepfakes in the 2026 Midterms: How to Spot a Fake Political Ad: the election deepfake landscape and what no federal law means for voters
- That Celebrity Crypto Video Is Probably a Deepfake: how financial fraud deepfakes are prosecuted under existing fraud statutes
- Three US Politicians Shared an AI Image as Real: The Iran Airman Incident: the liability and accountability questions when the people sharing AI fakes are elected officials themselves
