Google is expanding its AI detection system, SynthID, making features like Circle to Search and Lens capable of identifying AI-generated and AI-edited images. SynthID is a watermarking system that appends invisible metadata to content created or modified using Google’s AI tools, and the expansion includes Google Chrome and Search functionalities.
The updated system enables users on Android to utilize SynthID for assessing an image’s origin. Circle to Search can flag AI content, while Google Lens and the Chrome version of Gemini will provide answers to questions regarding whether certain images are AI-generated. An example highlighted by Google specifies that an image originated from a Pixel phone and was edited using AI tools within Google Photos.
However, Google acknowledges the limitation in providing detailed provenance for every AI-altered image. The company has integrated content credentials into its Pixel 10 series and plans to expand this feature to the Pixel 8 and 9 models soon. While Google can reliably flag images edited through its own tools, it may not accurately track content from other AI platforms.
Video: Google
OpenAI, Kakao, and ElevenLabs have committed to implementing SynthID technology for their AI-generated content. OpenAI stated its integration will begin with images produced by ChatGPT, Codex, or through its API. Google is also enhancing the support for content credentials, with the Gemini app receiving updates today and further integrations in Chrome and Search anticipated within the coming months.
Despite these advancements, Google noted that no AI watermarking system is infallible, and there exist techniques to bypass detection tools and watermarks. The company maintains that streamlining the verification process for AI-generated content represents an important initial measure as such content continues to proliferate.
