OpenAI is making it easier to check if an image was made by their models

The detail worth sitting with is that OpenAI is integrating SynthID, a watermarking technology developed and owned by Google. That means OpenAI is now dependent on a direct competitor's infrastructure for a core trust feature, which is an unusual concession that the announcement framing largely obscures.

The timing here is pointed. The same day this dropped, we covered Google's Flow platform lowering the barrier to synthetic selfie creation ('Google Makes It Easy to Deepfake Yourself'), Gemini Omni enabling conversational video manipulation, and Google restructuring Search around AI-native interfaces. Google is simultaneously making synthetic media easier to produce and supplying the authentication layer that other labs, including OpenAI, now depend on to prove provenance. That is a structurally advantageous position: Google shapes both the supply of synthetic content and the credibility infrastructure used to label it. OpenAI's adoption of SynthID accelerates Google's standard rather than a neutral one.

Watch whether Anthropic or Meta adopt SynthID within the next two quarters. If they do, Google's watermarking technology becomes the de facto industry standard before any regulatory mandate forces the question, which would give Google significant leverage over how provenance verification is governed going forward.