DEV Community

Cover image for AI Deepfakes: US Liability vs. EU Labels Clash
Tom Morgan
Tom Morgan

Posted on

AI Deepfakes: US Liability vs. EU Labels Clash

TL;DR The federal TAKE IT DOWN Act (signed May 19, 2025) criminalizes nonconsensual intimate imagery—AI-generated included—and forces “covered platforms”…

Article 50 of the AI Act, enforceable since August 2, 2026, doesn’t ban synthetic intimate content—it requires that it be disclosed as artificial, clearly, and at first contact. It makes it a federal crime to knowingly publish, or threaten to publish, a nonconsensual intimate image of an identifiable person—explicitly including AI-generated “digital forgeries,” not just real photographs. The Act criminalizes publication and distribution, not the act of an AI system producing the image in the first place.

Key takeaways

  • It has nothing to say about whether sexual deepfakes should exist.
  • Its entire theory of harm is that people are being deceived about what’s real, and its entire remedy is disclosure.
  • Article 50 can force a platform to caption a synthetic intimate image of a real person as artificial—and once labeled, the content is compliant.

If reliable adult verification and content classification were solved problems, “adult mode” would already exist.

Read the full article: https://www.forbiddenai.site/ai-deepfakes/

Top comments (0)