Are Deepfakes Illegal? US Deepfake Laws Full Guide

Are deepfakes illegal deepfake laws is deepfake porn illegal what states are deepfakes illegal TAKE IT DOWN Act DEFIANCE Act deepfake legislation 2025 2026 comes down to this: some deepfakes are clearly illegal (especially non-consensual sexual deepfakes), while others sit in legal gray areas (like non-sexual “fake audio/video” used to embarrass someone). In the U.S., the TAKE IT DOWN Act (signed May 19, 2025) makes non-consensual intimate images illegal even when they’re AI-generated, and it forces platforms to remove properly flagged content within 48 hours.

• What are deepfakes?
• How do deepfakes work (in plain English)?
• Why deepfakes matter in real life
• Real-world examples that shaped deepfake laws
• Are deepfakes illegal? Deepfake laws in 2025–2026
• Is deepfake porn illegal?
• What states are deepfakes illegal?
• Gaps and gray areas in deepfake legislation (2025–2026)
• What you can do if you’re targeted by a deepfake
• FAQ: quick answers about deepfake laws
• Conclusion: where deepfake laws are headed

What are deepfakes?

A deepfake is media (usually video, images, or audio) that’s been created or altered with AI to make it look like a real person said or did something they didn’t. The key point isn’t “editing” in general—it’s that the content can convincingly imitate a real person’s face or voice.

Deepfakes show up in several common forms:

• Sexual deepfakes / NCII (non-consensual intimate imagery), including pornographic images that never happened
• Election deepfakes meant to mislead voters
• Fraud deepfakes (like a fake CFO video call)
• Defamatory or harassing deepfakes that aim to humiliate or intimidate someone

One reason lawmakers have focused on sexual deepfakes is the scale: research in the provided context estimates 98% of all deepfakes online are NCII/pornographic in nature, and 98% of deepfake victims are women (with women targeted 4.5x more often than men).

How do deepfakes work (in plain English)?

Most deepfakes rely on AI models trained on lots of examples of real people’s faces or voices. The model learns patterns—how a person looks when they turn their head, how their mouth moves, or what their voice “signature” sounds like—and then generates new, fake media that imitates those patterns.

Practically speaking, the workflow often looks like this:

1. Collect source material (photos, video clips, voice recordings—often pulled from social media).
2. Train or prompt an AI tool to recreate the person’s face/voice (or to paste it onto someone else).
3. Generate and refine outputs until it looks “real enough.”
4. Publish or use it—for harassment, porn, political influence, or fraud.

That last step—publishing or using the deepfake to harm someone—is where many laws now focus, because it’s the harm (and the intent) that’s easier to define and prosecute than the underlying technology.

Why deepfakes matter in real life

Deepfakes aren’t just “internet pranks.” The research context shows measurable harm in three big areas: sexual abuse, fraud, and erosion of trust.

1) Sexual deepfakes are the main problem by volume

According to the research context, deepfakes online grew from about 500,000 files (2023) to a projection of 8 million (2025), and 98% are pornographic/NCII. That’s why you see so many laws and platform rules framed around intimate images and consent.

2) Deepfake fraud has become a repeatable business model

The context reports a 2,137% surge in deepfake fraud attempts over three years, with a new attack every 5 minutes in 2024. Reported North America losses are listed as $200M in Q1 2025 and $347M in Q2 2025.

A widely cited real-world case: in February 2024, engineering firm Arup (Hong Kong) lost $25 million when a finance worker was deceived by a deepfake video call posing as the CFO and colleagues—the context notes this as the largest confirmed deepfake fraud case.

3) It hits kids and schools, not just celebrities

The context includes a documented pattern of school impacts: 15 schools affected in one cluster incident in 2025. This matters for parents and educators because minors are uniquely vulnerable—and many laws increase penalties when victims are under 18.

Real-world examples that shaped deepfake laws

Lawmakers didn’t act in a vacuum. Several high-visibility incidents helped push deepfake legislation forward.

Taylor Swift deepfake porn (January 2024)

In January 2024, sexually explicit deepfake images of Taylor Swift were viewed 47 million+ times before removal, according to the research context. The same context notes this incident directly accelerated the TAKE IT DOWN Act—because it made the “removal speed” problem impossible to ignore.

First federal conviction under the TAKE IT DOWN Act (April 2026)

The context lists the first conviction in April 2026: James Strahler II (Ohio) pleaded guilty after using 24+ AI platforms to generate 700+ images of neighbors, including minors. That case is important because it shows the federal law isn’t just symbolic: it can be enforced.

Are deepfakes illegal? Deepfake laws in 2025–2026

In the U.S., deepfakes are not “one thing” legally. Whether a deepfake is illegal depends on what it depicts (sexual vs. non-sexual), who it targets (adult vs. minor), what the intent is (fraud/harassment vs. satire), and what jurisdiction applies (federal vs. state).

Key U.S. federal deepfake laws (and near-laws)

• TAKE IT DOWN Act (signed May 19, 2025): criminalizes the non-consensual publication of intimate images, including AI-generated deepfakes. Penalties are up to 2 years for adult victims and up to 3 years for minors. It also requires online platforms to remove flagged content within 48 hours.
• DEFIANCE Act (passed U.S. Senate unanimously January 13, 2026): creates a federal civil right to sue creators/distributors, with up to $150,000 in statutory damages (and $250,000 in aggravated cases). As of May 2026, the context states it is not yet law because the House has not voted.
• DEEPFAKES Accountability Act (H.R.5586) (introduced 2023 by Rep. Clarke): would require mandatory digital watermarking and disclosure on deepfake content, but the context says it stalled in committee and never passed.
• NO FAKES Act (H.R.2794) (reintroduced May 2025): would create the first federal right of publicity protecting an individual’s voice/image/likeness from AI replication. The context notes backers including SAG-AFTRA, Amazon, OpenAI. It is still pending.

Quick comparison: what the big federal bills do

• TAKE IT DOWN Act: criminal penalties + fast takedown requirement (48 hours) for non-consensual intimate images (real or AI).
• DEFIANCE Act: would let victims sue for statutory damages (not yet law as of May 2026).
• DEEPFAKES Accountability Act: would require watermarking/disclosure (stalled).
• NO FAKES Act: would protect your likeness/voice at the federal level (pending).

For readers who want primary sources, you can look up the relevant bills on Congress.gov (official bill text and status).

Is deepfake porn illegal?

Often yes—and increasingly so—especially when it’s non-consensual and depicts an identifiable person. This is the area where the law is clearest.

Federal rule now: TAKE IT DOWN Act (since May 19, 2025)

Under the research context, the TAKE IT DOWN Act makes it a crime to publish non-consensual intimate images, including AI-generated deepfake porn. It also creates a practical lever that victims care about: a 48-hour removal requirement for online platforms once content is properly flagged.

Penalties under the Act in the context:

• Up to 2 years for adult victims
• Up to 3 years for minors

State rules: most states have NCII/deepfake porn laws

The context states that by spring 2026:

• 46 states have enacted at least one deepfake-related law.
• About 45 states have laws covering sexually explicit deepfakes (NCII).

So if you’re asking “is deepfake porn illegal,” the real answer is: it’s very likely illegal, and the trend line in 2025–2026 is toward stronger victim remedies and faster takedowns.

What states are deepfakes illegal?

Because states write their laws differently, it’s more accurate to say: many states make certain kinds of deepfakes illegal (like pornographic or election-related deepfakes), rather than banning “deepfakes” across the board.

Here’s what the research context confirms as of spring 2026:

• 46 states have at least one deepfake-related law.
• 30 states have election-related deepfake protections.
• About 45 states cover sexually explicit deepfakes / NCII.

Examples of specific state deepfake laws (named in the context)

• Tennessee ELVIS Act (effective July 1, 2024): the first U.S. law specifically protecting musicians’ voices from AI cloning by adding “voice” to state personal rights protections.
• Pennsylvania Act 35 (signed July 7, 2025): criminalizes deepfakes created with fraudulent or injurious intent, with penalties up to 7 years (felony) if used to defraud, coerce, or steal.
• California AB 2839 (2024): targeted political deepfakes, but the context notes key provisions were struck down in August 2025 by a federal court on First Amendment/Section 230 grounds.

That California example is the big warning sign: courts can be skeptical of broad restrictions on political speech. The context sums it up: courts have generally been skeptical of broad bans on political deepfake speech.

Gaps and gray areas in deepfake legislation (2025–2026)

If you’re wondering why deepfakes still feel “everywhere” despite new laws, it’s because important gaps remain. The research context names several explicitly.

1) Non-sexual, non-election deepfakes often fall through the cracks

Many laws focus on sexual abuse (NCII) or election interference. The context notes that non-sexual, non-election deepfakes (defamatory fakes) are mostly not covered by NCII laws.

2) There’s no U.S. federal disclosure/watermark rule (yet)

The context says the DEEPFAKES Accountability Act (H.R.5586) would require watermarking and disclosure, but it never passed. So today, in the U.S., there is no federal mandatory watermarking requirement in force.

3) There’s no federal “right of publicity” for everyone (yet)

The context highlights that the NO FAKES Act would create a first federal right of publicity over your voice/image/likeness, but it’s still pending. That’s one reason state-by-state protections (like Tennessee’s ELVIS Act) matter so much right now.

4) Platform liability is still contested

The context flags that “deepfakes-as-a-service” platform liability remains contested. Practically, this affects whether victims can reach the person who uploaded the content (often anonymous), the tool used to generate it, and/or the platform hosting it.

What you can do if you’re targeted by a deepfake

If you (or your child) are dealing with a deepfake—especially sexual deepfakes—your priority is usually: preserve evidence, get it taken down quickly, and find the right reporting channel.

Step-by-step: practical actions you can take today

1. Document everything first. The context recommends saving screenshots and URLs before reporting. (Even a quick screenshot of the page and the account name can matter later.)
2. Use the federal takedown path if it’s intimate imagery. Under the TAKE IT DOWN Act (signed May 19, 2025), victims have a right to demand platform removal within 48 hours after properly flagging covered content.
3. If a minor is involved, report to NCMEC. The context advises reporting to NCMEC for minors.
4. If you’re an adult victim of NCII, contact a specialist org. The context recommends the Cyber Civil Rights Initiative for non-consensual intimate imagery help.
5. Track legal developments and lawsuits. The context suggests checking /ai-lawsuits/ for ongoing legal cases involving deepfake platforms.

If you’re a parent, teacher, or school admin

The research context notes a cluster incident affecting 15 schools in 2025. When deepfakes hit schools, the response is usually part safety issue and part evidence issue. Preserve evidence, involve guardians, and route minor-related reports through NCMEC.

You can also browse Ban the Bots resources designed for families at /parents/.

If you’re worried about deepfakes at work (fraud, HR, reputational harm)

The fraud numbers in the context (including the Arup $25M case and the 2024 “attack every 5 minutes” figure) show this isn’t hypothetical. A practical takeaway is to treat unexpected payment requests or “emergency” instructions on video calls as high risk, and require a second verification channel.

For broader accountability and policy steps ordinary people can support, see /fighting-back/ and /ai-backlash/.

FAQ: quick answers about deepfake laws

Are deepfakes illegal in the U.S.?

Some are. In the U.S., deepfakes tied to non-consensual intimate imagery are clearly targeted by federal law under the TAKE IT DOWN Act (signed May 19, 2025). Many states also regulate election deepfakes and sexually explicit deepfakes, but laws vary.

Is it illegal to make a deepfake if you never post it?

It depends on jurisdiction and the content. Internationally, the context notes the UK Data (Use and Access) Act 2025, Section 138, creates a crime for creating sexually explicit deepfake images without consent even if never shared (in force February 6, 2026). The U.S. approach described in the context emphasizes publication (TAKE IT DOWN Act) and state-specific rules.

What if the deepfake is “political speech” or satire?

This is where legal challenges happen. The context notes that California AB 2839 (political deepfakes) had key provisions struck down in August 2025 on First Amendment/Section 230 grounds, and that courts have been skeptical of broad bans on political deepfake speech.

Do platforms have to remove deepfakes?

For covered non-consensual intimate imagery, yes—under the context’s description of the TAKE IT DOWN Act, platforms must remove properly flagged content within 48 hours.

Do other countries require labels on deepfakes?

Yes, some do. The context says the EU AI Act Article 50 requires disclosure labeling for all AI-generated deepfakes, taking effect August 2026. China’s Deep Synthesis Provisions include mandatory labeling and require user consent before using biometric data (effective January 2023; updated September 2025).

For more background on detection and legal approaches, you may also want Ban the Bots explainers on /explainers/deepfakes, /explainers/deepfake-laws, and /explainers/how-to-spot-a-deepfake.

Conclusion: are deepfakes illegal deepfake laws in 2025–2026

The most honest answer to are deepfakes illegal deepfake laws is deepfake porn illegal what states are deepfakes illegal TAKE IT DOWN Act DEFIANCE Act deepfake legislation 2025 2026 is: many of the most harmful deepfakes are now clearly illegal—especially non-consensual sexual deepfakes—because the TAKE IT DOWN Act created federal criminal penalties and a 48-hour takedown requirement. But big gaps remain for non-sexual defamatory deepfakes, federal labeling/watermarking, and broad protection of your voice and likeness.

If you want to track where these harms are showing up and what people are doing about them, explore /ai-lawsuits/ (legal cases), /fighting-back/ (practical actions), /ai-backlash/ (public pushback), /data-center-map/ (AI’s physical footprint), and /ai-layoffs/ (how AI impacts jobs and communities).

External sources for deeper reading: the official bill tracker at Congress.gov; the National Center for Missing & Exploited Children (NCMEC) for minor-related reporting; and the Cyber Civil Rights Initiative for NCII support.