Google launches SynthID.com to check AI watermarks from partners

In short: Google has launched SynthID.com, a website that lets anyone check whether an image, video, or audio file includes a hidden SynthID AI watermark.

What happened

Google says content made with its Gemini AI models includes a hidden SynthID label. This label is like an invisible stamp that is baked into the media itself. For images and video, it is encoded into the pixels, and for audio it is encoded into the sound wave.

Until now, checking for that stamp was awkward for most people. If you were not part of Google’s trusted tester program, the main option was to ask Gemini to check, and the answer was not always a clear yes or no.

Starting today, people can use a dedicated site, SynthID.com, to do the check directly. Google says the updated detector can now identify SynthID watermarks from all of its partners, including OpenAI, Nvidia, and Kakao, and it expects Apple support soon. This matters because some partner-made AI content would not show up in Google’s older detector.

The website requires a login, and it supports Google, OpenAI, and Apple accounts. Google says users will have a daily limit of about 10 checks across images, video, and audio.

Why it matters

AI-made media can be used to mislead people, especially when it looks real. A single public checker makes it easier for journalists, teachers, and everyday users to quickly see whether some popular AI tools helped create or edit a file. Still, Google notes that not all AI systems use SynthID, and some use no watermark at all, so a “no watermark found” result does not prove something is real.

Source: Arstechnica