Sora now allows you to better control your AI versions, giving you greater ability to decide how and where your deepfake versions will appear in the app. This upgrade comes as OpenAI tries to show that it truly cares about its users' concerns, as a wave of AI content threatens to flood the internet.
See also: AI and content creation: Threat or opportunity for creators?
The new controls are part of a broader series of updates over the weekend aimed at stabilizing Sora and managing the chaos it creates in its feed. Sora is essentially “a TikTok for deepfakes,” a place to create 10-second of just about anything, including AI-generated versions of yourself or others (including voice). OpenAI calls these virtual appearances “cameos.” Critics call them a looming disinformation disaster.
Bill Peebles, head of the Sora team at OpenAI, said that users can now restrict how their AI-generated versions can be used in the app. For example, you can prevent your AI version from appearing in political videos, prevent it from saying certain words, or — if you hate mustard — prevent it from appearing near that condiment.
See also: OpenAI ChatGPT 4o: Testing a new watermark for the Image Generation model
OpenAI employee Thomas Dimsonsaid that users can also add preferences for their virtual versions, such as having them wear a hat that says “#1 Ketchup Fan” in every video.
Security safeguards are welcome, but the story of AI-powered bots like ChatGPT and Claude that provide advice on explosives, cybercrime or biological weapons suggests that someone, somewhere, will probably find a way to bypass them. People have already bypassed one of Sora's other security features, a weak watermark. Peebles said the company is also working to improve this.
Peebles mentioned that Sora will continueto “improve the restrictive capabilities” and “add new ways to stay in control” in the future.
See also: OpenAI launches Sora in the European Union
In the week since the app launched, Sora has helped to invigorate the internet with AI-generated content. The loose cameo controls — essentially a yes or no to groups like mutuals, approved by you, or “everyone” — have been a particular problem. The platform’s unwitting star, none other than OpenAI CEO Sam Altman, has been illustrating the danger, appearing in a variety of satirical videos that show him stealing, rapping, and even grilling a dead Pikachu.
