By collaborating with other companies, including its rivals, OpenAI wants to develop tools and technologies that allow viewers to trace AI-generated images, videos, and other content back to their source. This will help viewers avoid confusion between AI-generated content and real-world footage or photographs. The C2PA organization operates under the non-profit Joint Development Foundation and has released open-source technical standards for developers and companies to implement on their products. These standards enable the clear identification of content generated by AI models or other tools.
One of the standards developed by C2PA is the C2PA architecture, which stores and accesses cryptographically verifiable information. This architecture has already been used to create “Content Credentials,” a web-friendly watermark that provides information about who made an image, the tools used, and when it was created. OpenAI has been using the C2PA architecture to label images generated by its DALL-E 3 image generation AI model since February of this year. The architecture can also be integrated into the metadata of multimedia files, ensuring that even offline viewers can access information about the content’s origin.
OpenAI’s VP of Global Affairs, Anna Makanju, emphasizes the importance of these efforts in building trust in the authenticity of online content. OpenAI is also planning to integrate C2PA metadata into its video generating AI model called Sora, which is currently used by selected trusted partners. In addition, OpenAI has launched the DALL-E Detection Classifier Researcher Access Program, inviting researchers to test a binary classifier designed to predict whether an image originates from OpenAI’s DALL-E 3 model.
Furthermore, OpenAI, in collaboration with Microsoft, is launching a $2 million societal resilience fund. This fund aims to support AI education and understanding among older adults and individuals unfamiliar with the technology. The partnership will work with external groups like AARP, International IDEA, and Partnership on AI to achieve this goal.
The need for content provenance and authenticity has become more critical as social network sites like Facebook are seeing an increase in the sharing of AI-generated images that resemble real photographs. People are being tricked by these images, highlighting the importance of developing tools and standards to verify the authenticity of digital content. OpenAI’s involvement in C2PA and its initiatives demonstrate its commitment to responsible AI integration and ensuring that viewers can trust the content they encounter online.
