Google’s Gemini 2.0 Flash is known to bring a slew of experimental features, but not all of them are worth applauding. This lightweight AI model now supports native image production and conversational editing, helping users build and modify images effortlessly. Yet, over the weekend, a more problematic ability surfaced: The ability to perfectly erase watermarks from images.
While applications like Watermark Remover.io have been around for years, Gemini 2.0 Flash takes it up a notch, effortlessly erasing even sophisticated watermarks, including those from Getty Images. Here’s the twist: it substitutes the watermark with a SynthID mark, designating the image as “edited with AI.” But the funny thing here is that this AI can erase that mark, too.
Watermark removal isn’t the only red flag. Users have also noticed that Gemini 2.0 Flash can put recognised images of real individuals, such as Elon Musk, into photos—a capability its full-model equivalent explicitly prevents. This raises a can of ethical worms. Should AI possess such transforming power over images? And where are the measures to avoid abuse? For now, these features are restricted to developers via AI Studio, but the apparent lack of guardrails is troubling. We’ve asked Google for clarity, but so far, nothing.
With the recent news that Google and OpenAI have petitioned to the court to allow the use of copyright content to train their AI models, this feature clearly shows that AI will bring about a major shift in the way people create and release content on the internet. Yes, features like watermark removal have been around in online tools for years, but it is not the easiest to use. Now that Google Gemini can do it, the same feature becomes a lot more accessible and infinitely easier to use. Brands who make a living out of selling their photographs will obviously cry foul over this, but once the technology is out there, it will be picked up and replicated.
