Can You Remove a Deepfake If There's No Original Photo?

Yes. You can get a deepfake removed even if no real photo or video of the depicted event ever existed. The federal TAKE IT DOWN Act, like most state non-consensual intimate imagery (NCII) laws, doesn't ask whether the content copies something real. It asks whether the content is a non-consensual, realistic depiction of an identifiable person. A deepfake fabricated entirely from a handful of reference photos and a generative AI model qualifies just as a real leaked photo does, because the legal question is about consent and identity, not where the image came from.

Why this question even comes up

It's reasonable to assume that with no original, there's nothing to claim. That assumption is correct for one legal tool, the copyright-based takedown. It's wrong for the tool built for exactly this situation.

Why DMCA can't reach a pure deepfake

DMCA (17 U.S.C. §512) is a copyright mechanism. A valid DMCA notice generally has to assert rights over an actual copyrighted work: a photo or video that exists and was copied or distributed without permission. If a deepfake has no real photo or video behind it, because a model synthesized it from reference images instead of copying a specific real work, there often isn't a copyrighted work in the traditional sense to base a claim on. This is a structural gap in DMCA, not a flaw in how any individual notice is written. See does DMCA cover AI-generated content? for the fuller breakdown of where copyright law does and doesn't reach.

Why the TAKE IT DOWN Act closes that gap

The TAKE IT DOWN Act (Public Law 119-12, signed May 19, 2025) was written with this gap in mind. It covers non-consensual intimate imagery, explicitly including AI-generated "digital forgeries," the statute's own term for synthetic media depicting a real, identifiable person in a way a reasonable viewer would find indistinguishable from authentic. The depicted event doesn't have to have happened. The law asks about consent and realism, not provenance.

Most state-level NCII and deepfake laws that follow a similar model take the same approach: consent and identifiability are what matter, not whether the image is a "real" photograph.

What a valid notice needs

Under the TAKE IT DOWN Act, a compliant removal request to a covered platform needs four things in writing:

• A signature from the depicted individual, or someone authorized to act on their behalf.
• The precise URL(s), or other information sufficient to locate the specific content, not a general complaint about a site.
• A brief, good-faith statement that the content is non-consensual.
• Contact information sufficient for the platform to follow up.

A platform that qualifies as a "covered platform" under the statute must then remove the content, and make reasonable efforts to find and remove known identical copies, within 48 hours of receiving a complete, valid notice. An incomplete notice doesn't start that clock. It gets rejected or held until it's fixed, so getting the format right on the first submission matters. The TAKE IT DOWN Act eligibility checker walks through whether your situation qualifies before you file.

What you do and don't need to prove

With no original, the usual instinct is to look for proof of what the deepfake was made from. You don't need it.

• You don't need the source photo. There may not be one. A model can generate a believable image from a handful of public photos.
• You don't need to identify the AI tool, or to prove the technical process behind the fake.
• You do need the exact URL, a screenshot that shows the page and the account that posted it, and the date you found it. That is the evidence a platform acts on, and it is worth saving before you report anything, because reporting often removes the page.
• You do need to say the content is non-consensual, in your own words, in the notice. A short good-faith statement is enough.

If you do know which photo it was made from

Sometimes a fake is clearly built on a specific photo or video you took, or that you hold the rights to. In that case you may have a secondary DMCA claim over that underlying material alongside the TAKE IT DOWN Act notice. The notice for the synthetic content itself is still the more direct tool, since it doesn't depend on proving ownership of anything. Does DMCA cover AI-generated content? goes through where each applies.

Getting it out of search without an original

Search removal doesn't need an original either. Google has a separate process for involuntary fake explicit imagery, and a request can be filed alongside the takedown. See removing a deepfake from Google. Our de-indexing request generator helps you draft one.

Finding the copies

With no source file, there is also nothing to search for by filename. Finding copies means matching on the face, which is what our deepfake removal service does, or checking by hand using the steps in how to find out whether there is a deepfake of you. The full process for a video is in how to remove a deepfake video, step by step.

What this means in practice

If you're dealing with a deepfake that has no real photo or video behind it, don't assume you're out of options because a copyright claim doesn't fit. The TAKE IT DOWN Act (and comparable state laws) exists to cover this case, and its 48-hour compliance window for covered platforms is, in most cases, faster than DMCA's open-ended timeline. The practical difficulty is less about having a legal basis, which you do, and more about finding every instance of the content. A deepfake with no real source photo also has no "original" to trace copies back from. That's where identity-based detection, which matches directly against your face rather than a specific file, becomes the practical starting point whichever statute ends up applying. If you want to see what's already circulating, a free scan is a good place to start.

Part of our deepfake removal service: detection by face, notices filed for you, and monitoring for re-uploads.