OpenAI suspends Sora access
OpenAI has suspended early access to its text-to-video model, Sora, following direct backlash from artists. The suspension underscores persistent industry concerns regarding copyright, attribution, and the impact of generative AI on the creative labor economy. This move highlights the growing pressure on AI developers to address intellectual property and provenance issues in AI-generated content.
- The controversy escalated after OpenAI’s CTO, Mira Murati, was unable to specify whether the model was trained on videos from YouTube, a practice that YouTube CEO Neal Mohan has stated would be a “clear violation” of the platform’s terms of service. - Initially, OpenAI implemented an "opt-out" policy for its Sora 2 model, requiring copyright holders to request the removal of their intellectual property from the training data, a stance that drew significant criticism from Hollywood studios and artists. - In response to the backlash, OpenAI CEO Sam Altman reversed the policy to an "opt-in" model, giving creators more control over how their characters and content are used. The move came after users began generating videos featuring well-known intellectual property, such as characters from "South Park" and "Super Mario". - The second generation of the model, Sora 2, introduced significant advancements, including the ability to generate synchronized dialogue and sound effects, as well as more physically accurate and controllable video, which heightened concerns about its potential impact. - Legal challenges in the generative AI space are ongoing, with more than 25 pending lawsuits related to the use of copyrighted material for training large language models. Courts have issued conflicting rulings on whether using copyrighted works for AI training constitutes "fair use". - Beyond copyright, the release of Sora 2 raised fears about the potential for creating realistic deepfakes and disinformation, with experts noting that AI-generated videos of protests, arrests, and ballot fraud quickly appeared on the app after its release. - To address misuse, OpenAI implemented safety measures, including restricting prompts for violent or sexual content and adding a visible digital watermark to generated videos, though third-party tools to remove the watermark emerged shortly after release. - Access to Sora was offered through ChatGPT Plus and Pro subscriptions, with the Pro tier costing $200 per month for higher-resolution, watermark-free downloads, indicating a push towards professional creative workflows.
