The efficacy and vulnerability of Google DeepMind’s SynthID system are currently under scrutiny, following a software developer's claim to have reverse-engineered the AI watermarking technology. This developer asserts the ability to strip these digital tags from AI-generated images or, conversely, embed them into other works. Google, however, firmly refutes these assertions, creating a complex narrative around the security of its AI content identification.
The developer, known online as Aloshdenny, has publicly shared their methodology and code on GitHub. According to Aloshdenny, the process was surprisingly straightforward, requiring only 200 images generated by Google's Gemini AI, standard signal processing techniques, and, as he humorously noted, "way too much free time." He also informally mentioned that "a little weed also seemed to help" during his endeavors.
"No neural networks. No proprietary access," Aloshdenny stated on Medium, explaining his breakthrough. He elaborated, "Turns out if you’re unemployed and average enough ‘pure black’ AI-generated images, every nonzero pixel is literally just the watermark staring back at you." This suggests a fundamental discovery about the watermark's structure.
SynthID itself is an innovative, nearly invisible watermarking system designed by Google to tag content originating from its AI tools. It embeds directly into the pixels of images at the point of creation, engineered to resist removal without noticeable degradation of image quality. This technology is extensively deployed across Google’s AI ecosystem, marking output from models like Nano Banana and Veo 3, and even extending to AI-generated creator clones on YouTube.
Despite his claims of reverse-engineering, Aloshdenny expressed admiration for the system, describing it as "genuinely good engineering." He conceded that he was ultimately unable to completely remove SynthID in his tests. Instead, his approach focused on confusing SynthID decoders, making them fail to interpret watermarked images correctly rather than erasing the watermark itself.
The technical details of Aloshdenny’s method are intricate for those without a development background. A comprehensive breakdown is available on his Medium page, which he lightheartedly noted was "written up while Aloshdenny was 'high'."
Aloshdenny offered a critical perspective on the system's design: "The fact that the best I could pull off was confuse the decoder enough that it gives up — not actually delete the thing — says a lot about how well it was designed." He further clarified the system's practical goals, stating, "It’s not perfect. But it’s not trying to be unbreakable. It’s trying to raise the cost of misuse high enough that most people don’t bother.”
While the author of this report has not independently verified Aloshdenny’s project, current observations suggest that SynthID has not been definitively reverse-engineered to a degree that would allow for casual removal or insertion by novice users aiming to bypass AI detection systems. Google continues to dismiss the validity of Aloshdenny’s assertions.
"It is incorrect to say this tool can systematically remove SynthID watermarks," Google spokesperson Myriam Khan affirmed in a statement to The Verge. Khan emphasized the company's confidence in its technology, adding, "SynthID is a robust, effective watermarking tool for AI-generated content."
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.