SynthID

Google just dropped SynthID, an “invisible ink” watermark for AI-made stuff—images, audio, video, even text—baked in the instant it’s created and built to survive crops, filters, and compression. There’s even a Detector portal where you upload a file and it tells you if SynthID is lurking. Sounds neat… until the comments section lit up.

The hottest takes? Doubters are circling like hawks. One camp wants to know if this stamp can be dodged by simply re-generating the image with another AI. Another camp is already plotting stress tests: think CSI vibes—make a pure white image, make a pure black image, subtract the channels, hunt the hidden signature. Then there’s the existential drama: if these tools can only catch a subset, will people treat a negative result as “proof” it’s human-made? That’s how you get an “arbiter of truth”—and folks are not chill about that. A counter-movement demands the inverse: certify real camera photos so we can prove what’s authentic, not just what’s synthetic.

Meanwhile, receipts are being shared with links to Google’s own docs for the curious SynthID details. The memes? “Invisible tattoo,” “who watches the taggers,” and speedrunners racing to strip the mark. Transparency or tech theater? The internet’s still voting.