SynthID ships on by default, and default is the whole policy
Gemini Omni applies SynthID watermarking by default rather than as an option. For a provenance scheme, that single design choice determines almost everything about whether it works at population scale.
Optional provenance marking is close to useless, because the content you most want to identify is produced by exactly the people who would switch it off. Default-on inverts that: the marked set becomes everything except what somebody deliberately stripped, and stripping becomes an act with intent attached to it.
The scale argument follows from the distribution. Omni Flash reaches the Gemini app, Google Flow, YouTube Shorts and the YouTube Create app. A watermark applied by default across that surface covers a meaningful fraction of consumer video generation without any user making a decision — which is the only mechanism that has ever worked for a safety property at population scale.
The limits should be stated plainly. Watermarking is not tamper-proof; re-encoding, cropping and adversarial processing degrade it, and a determined actor removes it. It also says nothing about the many models that do not implement it, which is most of them, including the open-weight video systems now sitting near the top of the quality board.
So it is not a solution to synthetic media. It is infrastructure for the ordinary case — and it arrives as state legislatures write disclosure duties onto operators. A platform that can already prove which of its outputs were generated is holding the compliance artefact those rules will ask for.
The Next Web — Google launches Gemini Omni Flash with SynthID watermarking on by default → · Google AI for Developers — Gemini API release notes →