OpenAI Adopts Google's SynthID Watermark for AI Images with Verification Tool

Post by soraminazuki » Wed, May 20, 2026, 9:42 AM UTC

tux3 wrote: post #154

yorwba wrote: post #147

Earlier quoted context omitted.

You can use the verification tool to check whether the watermark is still detectable and iterate until you successfully removed it . Of course if you need to regenerate the image with an unwatermarked image-generation model to remove it, it more or less still serves its purpose.

How confident are you that the verification tool detects all of the watermark, and not just one layer? I would layer two watermarks and let the public remove the most visible one.

Post by spuz » Wed, May 20, 2026, 9:57 AM UTC

the_other wrote: post #151

hhh wrote: post #137

Earlier quoted context omitted.

Have you looked at twitter or Facebook and seen the swaths of our population that are just fully believing fake AI slop about politics, crime, etc?

Most of those people wont be checking the provenance of the images. FB have stopped from their fact-checking processes. Only the investigative or journalistically inclined will make use of this, and those people already fact check. Where’s the en mass gain?

Post by spuz » Wed, May 20, 2026, 10:06 AM UTC

WhatIsDukkha wrote: post #125

Barbing wrote: post #106

Earlier quoted context omitted.

Are there other politicians? Do disgruntled employees have bosses? Ex partners? A national news story in the US tonight, Lyft driver caught faking photos of his messy car. Not the most intelligent fraudster as he left the Gemini logo on the corner of the image. Providing these four examples in good faith :) also generally I _dislike_ DRM

Against MILLLIONS of legitimate uses which you don't seem to care to protect? You should also think about whether, suddenly, courts can now trust images they see because this technology exists? I think thats not even basically plausible.

Post by fc417fc802 » Wed, May 20, 2026, 10:48 AM UTC

bsder wrote: post #140

fc417fc802 wrote: post #117

Earlier quoted context omitted.

No, it wouldn't. ContentID is already used by Google for that exact purpose. They appear to be fully in favor of enforcing IP law provided the owning party raises a complaint.

ContentID is generated internally by Google and you have to qualify (aka be able to threaten Google with lawyers) in order to effectively participate. It also seems to be defeatable by such advanced techniques as flipping the video left to right. If SynthID works, this would allow people to tag their own videos with a watermark that is invariant across various levels of compression and editing. It would enable automa…

Post by rationalist » Wed, May 20, 2026, 10:57 AM UTC

WhatIsDukkha wrote: post #91

surgical_fire wrote: post #80

Earlier quoted context omitted.

If they don't matter, neither does a watermark.

and you need the watermark to tell you the fish with the mustache is fake... What image is going to change your worldview so radically that the drm saves you? edit - to be clear you are watermarking 100,000 fishes with mustaches because of your concern over 1 image that "matters" (and you don't even have an image that matters in mind)