Google rolls out improved SynthID AI content detector, now available globally

Key Points:

• Google’s SynthID, an invisible watermark embedded in AI-generated images, videos, and audio, is now detectable via a public website, SynthID.com, which supports watermarks from Google’s partners including OpenAI, Nvidia, Kakao, and soon Apple.
• Previously, only Google’s internal tools could verify SynthID, and only for Google’s own content, but the new site allows users to check AI-generated media from multiple sources with a simple yes/no result after logging in.
• Users have a daily quota of about 10 checks to prevent misuse, such as attempts to create SynthID removal tools, with support for login via Google, OpenAI, and Apple accounts.
• Despite the expanded detection, many AI models either use different watermarking methods or none at all, meaning the SynthID tool cannot identify all AI-generated content, and there is a risk SynthID marks could be removed in the future.
• While SynthID is currently the most effective tool for detecting AI fakes, experts suggest cryptographic labeling methods like C2PA might offer a more robust solution for authenticating original media going forward.