Identifying AI-generated images with SynthID

To identify AI-generated images created by Imagen, SynthID has developed a tool for watermarking and identifying AI-generated content, using two deep learning models optimized on a range of objectiveives to ensure accurate image identification. The combined approach includes SynthID contributing to the broad suite of approaches for identify images based on metadata. With this technology, creators can empower people with knowledge and information about when they’re interacting with generated media, and help prevent the spread of misinformation. Several ways were explored in building SynthID, including using text-based image recognition models to search for watermarks on images, and visual features that can be used for identification. The tool has been rigorously tested and proven robust, allowing for safe, secure, and trustworthy approaches at every step of the way. It can work with a broad range of AI models, making it available to third parties in Google Cloud and beyond. SynthID’s approach is being continuously evolved and optimized, with the goal of creating more effective solutions that empower people and organizations to responsibly work with AI-generated content.