OpenAI's Sora 2 video app raises copyright and deepfake concerns
OpenAI is developing Sora 2, a standalone AI video generation app with a TikTok-like interface, reportedly in early testing. The app enables realistic deepfake creation through a "Cameos" feature using biometric data, raising concerns about copyright infringement, identity misuse, and potential for misinformation despite OpenAI's built-in safety filters.
What happened
OpenAI is developing Sora 2, a standalone app for its video generation model with a vertical, swipe-to-scroll feed similar to TikTok where all content is AI-generated. Users can create AI-generated clips up to 10 seconds long, and the app allows identity verification so verified users can appear in videos through a 'Cameos' feature using biometric data uploads.
The app enables realistic deepfake creation, with numerous videos reportedly showing CEO Sam Altman in various scenarios. OpenAI began internal testing in late September 2025, and industry observers speculate a public release in late 2025.
Reported impact
Users have reportedly created videos featuring known personalities in unusual situations, raising concerns about personality rights violations. The deepfake capability and ability of users to circumvent safety measures create potential for misuse in misinformation and harmful applications.
OpenAI faces multiple existing lawsuits over alleged copyright violations, and Sora 2 includes filters to prevent videos that could infringe copyright, though the adequacy of these protections for the video generation model remains a concern.
Index analysis
The app's Cameos feature enabling users to create realistic deepfakes of both themselves and potentially others through biometric data raises significant risks of identity misuse and non-consensual deepfake creation. The reported ability of users to bypass safety measures suggests potential enforcement gaps in content moderation.
Sora 2 raises intellectual property concerns given OpenAI's existing copyright litigation, with questions about whether video generation models trained on existing media can comply with copyright protections despite built-in filters.
Governance controls
OpenAI has implemented filters to prevent videos that could infringe copyright and has added parental controls and is developing an age-prediction system to restrict users under 18 from certain interactions. Users are notified whenever their likeness is used, even in draft clips, according to disclosed features.
Despite these safeguards, no official age restrictions for Sora 2 have been disclosed, and users have reportedly found ways to circumvent existing security measures.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.