Generally, no. DMCA (17 U.S.C. §512) is a copyright enforcement mechanism, and copyright protects a specific original work — a photo, a video, a piece of writing that someone created and owns the rights to. A purely AI-generated deepfake, built from a model's synthesis of reference images rather than by copying an existing photo or video, usually isn't infringing anyone's copyright, because there typically isn't a copyrighted work being copied in the first place. You don't hold copyright in your own likeness the way you'd hold copyright in a photo you took or a video you filmed — likeness and copyright are legally different things, and DMCA only protects the second one.
Why "it's my face" isn't the same as "it's my copyrighted work"
This is the part that trips people up. It feels intuitive that content depicting you should be something you can control through copyright — but copyright law doesn't work that way. Copyright protects the expression a creator fixed in a work: the specific photograph, the specific footage, the specific arrangement of pixels someone authored. It doesn't protect a person's face, name, or identity as a general matter — that's a separate area of law (right of publicity, and for non-consensual intimate content specifically, statutes built for exactly that purpose).
So when a deepfake is generated purely from a model's own synthesis — no specific real photo or video was copied to make it — there's typically no underlying copyrighted work for a DMCA notice to point to. The person depicted didn't author the deepfake, and the deepfake wasn't made by copying something they did author.
The narrow exception
There's a real edge case worth knowing about: if a deepfake was built by directly feeding in the depicted person's own copyrighted photo or video — for instance, a generative "face-swap" or editing process applied to a specific real image the person took or owns — DMCA might reach the underlying act of copying that source material, even if the final output is heavily transformed. That's a fact-specific legal question (courts still work through exactly how transformative use and AI training interact with copyright), and it depends entirely on whether an identifiable, owned original was actually used as an input — not on the general fact that AI was involved.
In practice, most of the deepfake content that harms people isn't built this way. It's generated from a broader model trained on many images, producing new synthetic output rather than directly transforming one specific copyrighted photo the person can point to.
Don't rely on a DMCA notice as your only tool
If you file a DMCA notice against a pure AI fabrication with no traceable underlying copyrighted work, a platform reviewing it for validity may reject it — not because the harm isn't real, but because the notice doesn't meet copyright's specific requirements. Using the wrong legal tool can cost you time you don't have.
What actually covers pure-likeness deepfakes
This is exactly the gap the federal TAKE IT DOWN Act (Public Law 119-12, signed May 19, 2025) was built to close. It covers non-consensual intimate imagery including AI-generated "digital forgeries," and the statute explicitly does not require that the depicted event ever really happened — the operative question is whether the content is a non-consensual, realistic depiction of an identifiable person, not whether it copies a real prior work. Covered platforms must remove qualifying content within 48 hours of a complete, valid notice. See can you remove a deepfake if there's no original photo? for the full mechanics of that process.
State-level right-of-publicity laws are the other relevant tool, particularly for deepfakes that aren't intimate imagery specifically (e.g. a deepfake used for a fraudulent endorsement) — these vary considerably by state, but generally protect a person's right to control commercial or otherwise unauthorized use of their identity, independent of copyright entirely.
DMCA vs. TAKE IT DOWN Act, side by side
| DMCA | TAKE IT DOWN Act | |
|---|---|---|
| Legal basis | Copyright ownership of a specific work | Non-consensual depiction of an identifiable person |
| Requires a real underlying photo/video? | Effectively yes | No |
| Covers pure AI-generated deepfakes? | Generally no | Yes, explicitly ("digital forgeries") |
| Response deadline | None fixed by statute | 48 hours for a complete, valid notice |
| Who can file | The copyright holder | The depicted individual, or an authorized representative |
The bottom line
DMCA is the right tool when real copyrighted footage of you was taken and shared without permission — it is generally the wrong tool for a deepfake fabricated without a traceable copyrighted source, because copyright protects specific works, not likeness itself. The narrow exception is a deepfake built by directly manipulating a specific copyrighted photo or video you own, which may still trigger a copyright claim over that underlying material. For everything else — the more common case of a purely AI-generated depiction — the TAKE IT DOWN Act and state right-of-publicity law are the mechanisms actually designed to reach it. Filing the wrong notice type doesn't just fail, it can cost you the time a faster, correctly-targeted notice would have saved. If you're not sure which applies to something you've found, a free scan is a good starting point to see what's out there before deciding how to act on it.