\
Insight
\
\
Insight
\
What is deepfake pornography? The harm, the law and how platforms detect it
Gabe Regan
VP of Human Engagement
Published June 1, 2024 Updated October 20, 2026
Deepfake pornography is sexually explicit imagery, video or audio created with AI to show a real person without their consent. It’s a form of image-based sexual abuse. Most victims are women. In the U.S., the TAKE IT DOWN Act requires covered platforms to remove reported nonconsensual intimate imagery, including deepfakes, within 48 hours.
A woman opens a link a coworker forwarded and finds her own face on explicit images she never posed for. She didn’t consent, and she can’t tell how far the files have already spread. That moment is what deepfake pornography does to the people it targets. It also leaves platforms with a hard question about what’s real and how fast they can act.
Who deepfake pornography targets
The numbers are lopsided. Security Hero’s 2023 State of Deepfakes report analyzed 95,820 deepfake videos and found that 98% were pornographic. In that material, 99% of the people targeted were women.
Most victims never agreed to any of it, and many don’t know it exists until someone else finds it. The imagery doesn’t need a celebrity. Face-swap and so-called nudify apps turn ordinary photos from social media into explicit content, so anyone with a public image can become a target.
The harm doesn’t end when a file comes down. Copies move to other sites, and the person in them carries the reputational and emotional cost long after. Reality Defender has written about what we owe to victims of nonconsensual deepfakes.
Why the abuse spreads across formats
Nonconsensual synthetic imagery rarely stays in one format. The same likeness can turn up as a still image, a video and a cloned voice attached to either one. A moderation queue that screens images alone sees one surface of the same abuse.
That’s why trust and safety teams ask about coverage before they ask about accuracy. A takedown is only as complete as the formats the team can check. Identity is now synthetic, and the abuse follows the person across every format their likeness can take.
What the law requires of platforms
Federal law changed in 2025. The TAKE IT DOWN Act, signed May 19, 2025, requires covered platforms to remove reported nonconsensual intimate imagery, explicitly including AI-generated deepfakes, within 48 hours of a valid report. That platform obligation has applied since May 19, 2026, and the FTC enforces it (Congressional Research Service).
The DEFIANCE Act would let victims sue the people who create or knowingly share sexual deepfakes of them. The Senate passed it unanimously in January 2026, as covered in Reality Defender’s explainer on the DEFIANCE Act. State laws vary widely, and Public Citizen tracks them state by state.
Outside the U.S., the UK has made creating a sexually explicit deepfake without consent a criminal offense. Reality Defender’s breakdown of the UK legislation covers what that change means for platforms.
Where detection fits in a takedown
A 48-hour clock changes how platforms work. Teams need to confirm that reported content is AI-generated, find copies of it and act before the deadline passes. Doing that by eye doesn’t scale.
Detection gives moderators a score for whether an image, video or audio file shows signs of AI manipulation. It can run at upload, before content goes public, or on the reports queue, where it helps reviewers prioritize. Reality Defender’s RealScan analyzes images, video and audio from one platform, so a team checks every format a report arrives in.
Detection doesn’t replace human review or a victim’s report. It gives the people making takedown decisions a faster, consistent signal to work from.
Frequently Asked Questions
What trust and safety teams ask about deepfake pornography
Creating and sharing it is restricted by a mix of federal and state law. The federal TAKE IT DOWN Act makes it a crime to knowingly publish nonconsensual intimate imagery, including AI-generated deepfakes, and requires covered platforms to remove reported material within 48 hours. Many states have their own laws, and the details differ by state.
Since May 19, 2026, covered platforms must remove reported nonconsensual intimate imagery, including AI deepfakes, within 48 hours of a valid report. The FTC enforces the requirement. The law covers intimate imagery specifically, not deepfakes in general.
Report it to the platform hosting it, since covered platforms must act on valid reports within 48 hours in the U.S. Adults can use StopNCII.org to create a digital fingerprint of the image that participating platforms use to block it. For anyone under 18, NCMEC’s Take It Down service does the same.
Detection models look for signs of AI generation in the file itself, such as artifacts in pixels, motion or audio that real recordings don’t carry. The result is a score that helps reviewers decide what to act on first. It works alongside human review and victim reports rather than replacing them.
Deepfake pornography harms the person in the image first and the platform that hosts it second. The law puts a 48-hour clock on the response, and detection is how teams meet it at scale. Detection is a continuous discipline, and platforms that treat it that way protect the people their users can see.onomy and takes concrete steps to prevent the exploitation and harm caused by this type of content.
\
Insights