A federal law sets a 48-hour deepfake takedown clock for platforms

The classroom note

A parent hears that a fake, sexually explicit image of a minor made with an AI tool is now a federal crime, and asks a school what that means in practice. The answer starts with a single piece of vocabulary the statute itself supplies. The TAKE IT DOWN Act, signed into law on 19 May 2025 as Public Law 119-12, defines a digital forgery as any intimate depiction of an identifiable person created through software, machine learning or similar computer-generated means that, viewed as a whole by a reasonable person, is indistinguishable from an authentic depiction. That definition, not the word deepfake, is what the statute's criminal and takedown provisions actually turn on.

What the evidence says

The statute's text, read alongside the Congressional Research Service summary attached to the bill's record, sets out two mechanisms. First, it is a federal crime to knowingly publish an intimate depiction, authentic or a digital forgery, of an identifiable minor with intent to abuse, humiliate or gratify sexual desire, or of an adult where publication lacks consent and is intended to or does cause harm; penalties reach three years' imprisonment for offences involving minors. Second, separately, the Act requires covered platforms, broadly, sites and apps that primarily host user-generated content, to build a notice-and-removal process within one year of enactment, by 19 May 2026, and to remove a validly reported depiction within 48 hours, enforced by the FTC as an unfair-or-deceptive-practice matter.

The implementation question

A school is not itself a covered platform under the Act, so its role sits beside the statute rather than inside it: recognising that a depiction reported to it, if hosted elsewhere online, has a specific, dated legal removal channel a family can use, and that the criminal provisions apply however the image was produced. The one-year runway before the removal duty takes effect, distinct from the law's immediate criminal provisions, is worth naming plainly to families expecting an instant fix, since the platform-side machinery the Act requires was not obliged to exist until roughly a year after signing.

What holds and what fails

The statute's protection holds cleanly for the case it was written around: a specific, identifiable person depicted without consent, reported through a platform's required process. It says nothing about school disciplinary procedure and, being a federal criminal statute, cannot resolve a case without a report to the authorities it names. Reading the Act as a general AI-image law overstates what its text, focused specifically on intimate depictions, actually covers.

• Is the case in question an intimate depiction as the statute defines it, or a different kind of manipulated image the Act does not reach?
• Has the one-year platform deadline of 19 May 2026 already passed for the service where the image was posted?
• Does the family need the platform's removal process, a law-enforcement referral, or both?

The Act gives a name and a clock to one specific harm; it is not, by itself, a school's incident-response plan.