Particle.news

Download on the App Store

OpenAI Revises Sora 2 Policy, Pledges Rights Controls and Revenue Sharing After Viral Deepfakes

OpenAI says rights holders will get more granular control over character generation as high-fidelity Sora 2 clips exposed copyright and impersonation risks.

Overview

  • Sora 2 launched last week as an invite-only, TikTok-style text-to-video app in the U.S. and Canada, with videos currently capped at about 10 seconds.
  • Sam Altman said OpenAI will let rights holders block or specify use of their characters and will test a revenue-sharing model for those who grant permission.
  • Early outputs included widely shared clips such as SpongeBob in a meth-lab parody and a fake Sam Altman shoplifting video, underscoring deepfake and IP concerns.
  • WME told agents it has opted out all clients from the latest Sora update, and Disney was reported to have opted out as studios weigh legal and licensing exposure.
  • Demos show major gains in realism, lighting, and scene control alongside persistent physics and micro-interaction flaws, and OpenAI’s filters and takedown tools have not prevented problematic generations.