SynthID
Google DeepMind's invisible watermark, hidden inside AI-generated images, audio or video rather than attached as a label.
SynthID marks AI-generated content by changing the content itself in ways people cannot see or hear. That is the key difference from ordinary metadata, which sits alongside a file and disappears the moment someone re-encodes it, screenshots it, or sends it through a messaging app. A SynthID mark is designed to survive cropping, filters and compression.
It was built by Google DeepMind and is now also used by OpenAI, which embeds it in audio from its GPT-Live voice model. Worth knowing the limits: a detected watermark reliably tells you the file came from that generator, but no watermark tells you nothing, since plenty of AI tools do not mark their output at all.
-
Google's New Voice Models Keep Talking While They Think
-
Google Switches Off Three Imagen 4 Models Today
-
Everything Claude Writes Now Carries a Hidden Mark, Everywhere in the World
-
Google put an image generator in Google Earth, then pulled it a day later
-
OpenAI is now watermarking AI voices, and you can check a file yourself