OpenAI appears to be walking back its Sora copyright policy (www.businessinsider.com)

OpenAI’s new text-to-video app Sora 2 quickly filled with clips of well-known characters after its launch, prompting legal and industry blowback over potential copyright infringement. Reports said OpenAI initially planned an “opt-out” approach—rights holders would have to ask not to have their characters generated—but CEO Sam Altman now says the company will give rights holders “more granular control” over character generation, similar to an opt-in/likeness model, and explore revenue-sharing for creators who permit use of their IP. For the AI/ML community this matters as both a technical and policy precedent: platforms that generate audio/video from text prompts must implement identity/likeness controls, metadata and filtering systems, and potentially negotiate licensing or revenue-split mechanisms. Sora’s case echoes prior disputes over voice and stylistic imitation (e.g., Studio Ghibli–style images, contested chatbot voices) and sits alongside ongoing lawsuits against OpenAI. How OpenAI engineers enforcement (automatic detection, access restrictions, opt-in APIs, or paid licensing) will influence industry norms for generative models, developer tooling, and commercial viability for text-to-video services.