What changed
Google has launched SynthID.com, a global public detector for invisible watermarks in AI-generated or edited images, video and audio. The service combines Google’s SynthID checks with partner systems from OpenAI, Nvidia and Kakao, with Apple support planned; users log in with a Google, OpenAI or Apple account and get roughly 10 checks per day.
Google says SynthID already covers more than 180 billion Gemini images and videos, plus 240,000 years of audio. The public tool gives a simple result: whether it found a supported SynthID watermark. It does not show which parts of an image contain the signal.
Why it matters
This turns a previously awkward check into a single web destination. Someone investigating a suspicious image, video or audio file no longer has to ask Gemini and interpret a vague answer. They can upload the file, wait for a yes-or-no result, and move on.
That is useful, but it is not an authenticity machine. A clean result means only that the tool did not find one of the supported SynthID watermarks. Content made by non-partner systems, tools that use no watermark, open-weight models, or systems that remove the signal may pass through undetected.
The daily limit is deliberate. Google engineers say unrestricted checks could help people develop tools to strip SynthID, and repeated tests of nearly identical files may trigger an earlier lockout. For technology teams, that makes SynthID a practical screening layer, not a final verdict: a positive result is meaningful, while a negative result leaves plenty of room for doubt.
Apple’s planned support could broaden the service further, but its launch timing and detection accuracy are not yet established.
Comments
No comments yet.