OpenAI has discontinued public access to its Sora text-to-video generation model, the company confirmed. The AI tool, which was released for limited testing in early 2025, could create highly realistic, minute-long videos from text prompts.
The decision follows widespread concern from researchers and policymakers about the potential for such technology to generate convincing non-consensual deepfakes and disinformation. OpenAI stated that the current version of Sora presented "unacceptable risks" for public deployment without more robust safety measures.
While the model demonstrated significant technical advancement, its public availability was paused indefinitely. The company indicated it would continue internal research and development, focusing on implementing stronger content provenance and detection systems before considering any future release.
The move highlights the growing industry and regulatory challenges surrounding advanced generative AI. Several jurisdictions are currently drafting legislation specifically targeting the creation and distribution of synthetic media.
