Google’s AI-detector for images and videos, SynthID, is getting a full release today. It will work for more than just content developed with Google’s AI models and initial partners.
SynthID is a necessary tool as AI content has taken hold of heavy portions of the internet. Models have evolved so much that it’s nearly impossible to tell whether some videos or photos are even real, and it has you second-guessing whether real ones are generated. SynthID takes aim at that problem by introducing invisible watermarks at the generation stage, and when used on an image, it will detect those watermarks.
Of course, that only works if the image generation tool inserts its own footprint to begin with.
Google announced that SynthID is being upgraded from limited early access to a globally available tool in English, though it’s expected to expand further in the future. The tool has progressed some, but this is a bigger step for AI content detection. According to Google’s press release, the AI image detector launch comes with expanded compatibility.
The new SynthID will work with content generated on Google’s own Gemini, a given, as well as content from more partners. So far, the list includes OpenAI, Nvidia, Kakao, and now, Apple. That means any content generated with models from those companies will contain visually undetectable watermarks so users can later check whether the image is a human-made visual or something an algorithm cobbled together.
Google notes that images made with Apple Intelligence will carry a SynthID watermark “soon,” which implies it might be a little bit before users can reliably identify AI-generated content from Apple’s models. But the expansion means that more users in total will be able to do some due diligence on photos and videos posted online.
FTC: We use income earning auto affiliate links. More.
Comments