A software developer says they figured out how to remove Google’s invisible watermarks from AI-generated images. The watermarks are supposed to help people identify when pictures were made by artificial intelligence.
This could be a big problem. Google’s SynthID system was designed to be tamper-proof, adding invisible markers to AI images that would stay there even if someone edited the photo. The whole point was helping people spot fake images in a world where AI can create incredibly realistic pictures.
The Cat and Mouse Game Begins
The developer, using the name Aloshdenny, posted their method on GitHub for anyone to use. They claim you can not only strip these watermarks away but also add fake watermarks to regular photos to make them look AI-generated.
Google says this isn’t true and that their system hasn’t actually been cracked. But the fact that someone is even attempting this shows the challenge tech companies face. Every time they build a new security system, someone tries to break it.
The timing is especially tricky. As AI-generated images become harder to distinguish from real photos, watermarking seemed like a promising solution. Major AI companies have been working on similar systems to help combat misinformation and deepfakes.
What Happens Next
This is likely just the beginning of an ongoing battle between AI companies and people trying to circumvent their safety measures. Even if this particular claim isn’t real, it won’t be long before someone does crack these systems. The question is whether watermarking technology can evolve fast enough to stay ahead of the hackers.
