OpenAI is launching the Sora app, its own TikTok competitor, alongside the Sora 2 model (techcrunch.com)

OpenAI today unveiled Sora 2, an upgraded audio-and-video generative model, alongside a companion social app called Sora that aims to compete with TikTok-style short-video feeds. Sora 2 is pitched as a more physically plausible video generator — OpenAI shows examples where objects respect motion and collisions (e.g., a missed basketball rebounding off the backboard instead of teleporting to the hoop). The app includes a “cameos” feature: users upload a one-time video+audio verification to capture their likeness, then can drop themselves (or grant friends permission to do so) into AI-generated scenes. Sora is available on iOS in the U.S. and Canada in an invite-only rollout; ChatGPT Pro users will get access to Sora 2 Pro without an invite. OpenAI says the app is free at launch, with pay-for extra generations during high demand. The launch matters because it pairs state-of-the-art generative video with a social distribution layer, raising fast questions for creators, platforms and regulators. Technical implications include stronger physical consistency in generative models and tight integration with user identity data (cameos + optional use of ChatGPT history for personalization). That integration enables richer, personalized feeds but heightens safety and privacy risks: non-consensual deepfake-style abuse, misuse by friends given cameo permission, content moderation challenges, and data‑driven recommendation concerns — all areas where OpenAI’s safeguards (revocable likeness access, parental controls, opt-outs) will be closely scrutinized.