NO FAKES Act of 2026 (US S.4591)

The NO FAKES Act would create a new federal property right protecting every person's voice and visual likeness from unauthorized AI-generated "digital replicas" — realistic, computer-generated recreations of a specific person in content where they didn't actually appear (or where their real performance was materially altered).

Anyone who creates an AI-generated recreation of a person’s voice or appearance without authorization could be sued. Licenses allowing use of a person’s likeness must be in writing, describe the intended uses specifically, and can't exceed 10 years while the person is alive. These rules apply to contracts signed after the law takes effect; existing contracts are grandfathered.

Platforms that host user-uploaded content can be sued if they receive a valid takedown notice and fail to act. To qualify for a safe harbor that limits their liability, similar to how the DMCA works for copyright, platforms must:

• Register a designated agent with the Copyright Office
• Have a policy for terminating repeat infringers
• Remove flagged content promptly when they receive a valid takedown notice
• Use digital fingerprinting to prevent re-uploads of content that's already been flagged

Platforms that do all of this face a maximum of $25,000 per work if they're found liable. Platforms that don't face up to $750,000 per work.

AI tools or services that are primarily designed to generate replicas of specific real people without their authorization are explicitly covered by the bill's liability provisions and don't qualify for the platform safe harbor. Any AI companion or content generation service using individuals' likenesses needs clear, written authorization.

Note: Counter-notifications disputing wrongful takedowns require a notarized physical signature — a higher bar than the DMCA's electronic signature standard, which could make it harder to fight back against bad-faith takedown requests.