Take It Down Act requires platforms to remove AI deepfakes within 48 hours under FTC enforcement
Platforms that primarily host user-generated content must now accept removal requests for AI-generated non-consensual intimate imagery and comply within 48 hours, or face FTC civil fines of approximately $53,088 per violation. The notice-and-removal enforcement regime became effective May 19, 2026, one year after the law was signed by President Trump.
- Issuing body
- U.S. Congress (statute); Federal Trade Commission (enforcement authority)
- Jurisdiction
- United States
- Who is bound
- Any website, app, social media service, video or image host, messaging app, or gaming platform that primarily hosts user-generated content — collectively defined in the Act as "covered platforms."
- Decided
- Effective
What's now different
Section 3 of the Take It Down Act, the provision the FTC enforces, requires covered platforms to establish a removal request mechanism accessible without a platform account, remove reported images or videos — and all known identical copies — across all storage infrastructure within 48 hours of a valid request, display clear notice of the removal process, and assign each request a tracking number. The 48-hour window applies to mirrored servers, CDN caches, and backup systems, not just the originally reported instance. Failure to comply is an unfair or deceptive act under Section 18(a)(1)(B) of the FTC Act, exposing platforms to the FTC's standard civil penalty ceiling — approximately $53,088 per violation as of May 2026. The FTC published compliance guidance on the enforcement date, recommending that platforms use perceptual-similarity hashing and share hash values with the National Center for Missing and Exploited Children's Take It Down service (for minors) and StopNCII.org (for adults).
The compliance obligation is technically harder than it appears. AI-generated images may never exist as a stored file until rendered on demand, making byte-level hashing insufficient — platforms need perceptual-similarity matching to catch near-identical re-uploads. For platforms without dedicated trust-and-safety engineering — smaller video hosts, community image boards, gaming services — the infrastructure represents a significant new build obligation; the law does not distinguish by platform size. The law covers "digital forgeries" — a statutory term for AI-generated intimate imagery — meaning AI image and video generation services that are themselves covered platforms face direct liability. The first federal criminal conviction under the Act (which took effect immediately on signing in May 2025) was handed down in April 2026 against an Ohio man who used AI to create non-consensual intimate imagery of neighbors, including children.