Google DeepMind launches an AI watermarking tool for identifying AI-generated images

SynthID is designed to identify AI-generated content.

Google DeepMind has partnered with Google Cloud to launch SynthID, a tool for watermarking and identifying artificial intelligence-generated images. The tool creates a digital watermark within the pixels of an image that is imperceptible to the human eye but detectable by an algorithm. SynthID is designed to identify AI-generated content.  Its creators claim it is immune to modifications such as filters and compressions. The tool is only compatible with Imagen, Google's text-to-image model available on VertexAI. DeepMind said SynthID would empower individuals with knowledge of when they are interacting with generated media, which could help prevent the spread of misinformation.

The technology is not foolproof but is a promising technical approach, and it could also evolve alongside other AI models and modalities beyond imagery, such as audio, video, and text. Other firms, such as Steg.AI and Imatag, have developed similar watermarking AI models. However, a common watermarking standard has been difficult to establish. A recent regulation issued by China’s Cyberspace Administration requires generative AI vendors to mark generated content without affecting user usage, while a U.S. senator has called for greater transparency in generative AI, including the use of watermarks.