OpenAI, best known for its ChatGPT platform, is making its most ambitious push yet into social media and entertainment. The company introduced Sora 2, its upgraded model for creating audio and video, and paired it with the launch of Sora, a new app that blends AI content creation with the short-form video feeds popularised by TikTok and Instagram Reels.
The dual rollout signals OpenAI’s intent to move beyond being a background technology provider and into the public-facing world of digital communities. Through Sora, users can generate clips of themselves and their friends in virtually any scenario—whether realistic or fantastical—and publish them directly to a central feed designed to encourage scrolling and discovery.
One of the app’s flagship tools, called Cameos, lets people train the system on their likeness with a single video and audio sample. After verification, their digital persona can be dropped into AI-created scenes: singing on stage, starring in a film sequence, or performing slapstick comedy. OpenAI says users can withdraw this permission at any time, but experts caution that the ability to mimic someone so convincingly raises the risk of misuse.
The app also leans heavily on data-driven recommendations. By default, Sora’s algorithm draws from a user’s in-app activity, their location, past interactions, and even their ChatGPT history to personalise what appears in the feed. A setting exists to disable this integration, and parents will have access to additional controls to limit scrolling, restrict recommendations, and manage who can contact their children.
Unlike traditional platforms that rely on advertising from day one, Sora will be free to use at launch, with revenue instead tied to how many videos people generate. During high-demand windows, users may need to pay to produce extra clips, a model that OpenAI believes will avoid putting too much pressure on its fledgling community.
Still, the arrival of a new AI-powered social network highlights unresolved issues around safety. Non-consensual videos remain a persistent problem online, and while OpenAI promises mechanisms to protect identities, the sheer ease of creating synthetic footage could make Sora fertile ground for abuse. Legal frameworks have not yet caught up to such scenarios, leaving much of the responsibility on the platform itself.
