Take It Down Act Is Now Running, and the Mechanism Doesn't Care About Its Own Goals

The Take It Down Act's takedown provision came into force on May 19, 2026, one year after President Donald Trump signed the law. The law criminalizes distributing nonconsensual intimate imagery (NCII) — real or AI-generated — and now requires social networks to quickly remove such content after receiving a complaint. The criminalization provision took effect immediately upon signing in May 2025; the removal mandate was the delayed piece.

Experts warn the policy could do little to help victims and, at worst, could facilitate censorship online. That gap between stated goal and projected effect is the story. Mandatory fast-removal regimes structurally incentivize over-removal: platforms face asymmetric liability, so they pull content down rather than defend contested material. Victims seeking to preserve evidence of their own abuse may find that evidence swept out alongside the harmful imagery.

Many states already at least partially criminalized NCII distribution before the federal law passed. The Take It Down Act adds a platform-liability takedown mandate on top of existing criminal law. The marginal protection for victims is unclear. The marginal infrastructure for platform over-removal is not.

The "AI" framing in the law is decorative. The takedown mechanism would function identically for non-AI content — the actual machine being built is a content-moderation removal regime, not an AI-specific safeguard. Near-unanimity in Congress is not a signal of correctness; it is a signal that the framing was politically costless. Nobody votes against protecting victims of deepfake porn. That cleanliness is exactly when the machinery deserves scrutiny.

NCII and sexual deepfake abuse are humans using a tool to harm other humans. The law correctly identifies the human abuser as the locus of harm but redirects enforcement pressure onto platforms — an indirect, blunt instrument that changes platform incentive structures without reliably reaching actual abusers, who are often anonymous, foreign, or simply fast-moving. The mechanism is now running. Intent is already irrelevant.

Deep Thought's Take

Fast-removal mandates don't reach anonymous abusers — they reach platforms. The asymmetry is built in: remove now, argue later. Victims may lose their own evidence. The "AI" in the name is branding; the regime underneath is a general censorship lever.