Users on social media have found a controversial application for Google’s Gemini AI model, specifically its ability to remove watermarks from images, including those published by Getty Images and other stock media providers.
Stock image companies are not going to like this
Last week, Google expanded access to its Gemini 2.0 Flash model’s image generation feature, allowing users to natively generate and edit image content. While this feature is powerful, it also lacks sufficient restrictions. Gemini 2.0 Flash can create images depicting celebrities and copyrighted characters and, notably, remove watermarks from existing photos.
Several users on X and Reddit have observed that Gemini 2.0 Flash not only removes watermarks but also attempts to fill in any gaps left by the removal. While other AI tools can do this, Gemini 2.0 Flash reportedly excels at the task and is available for free use.
It should be noted that Gemini 2.0 Flash’s image generation feature is currently labeled as “experimental” and “not for production use,” and is only accessible through Google’s developer-facing tools like AI Studio. The model does have limitations, as it struggles with certain semi-transparent watermarks and those that cover large portions of images.
Images: xXLeoXxOne on Reddit
The lack of usage restrictions in Gemini 2.0 Flash may raise concerns among copyright holders. Unlike Gemini, models such as Anthropic’s Claude 3.7 Sonnet and OpenAI’s GPT-4o explicitly refuse to remove watermarks. Claude describes watermark removal as “unethical and potentially illegal.”
According to U.S. copyright law, removing a watermark without the original owner’s consent is considered illegal, except for rare exceptions. Most AI image generation companies watermark their images to denote their origin, yet an unexpected application of these generators has emerged: removing existing watermarks from stock image repositories, where Gemini 2.0 Flash appears to be particularly adept.
Examples of watermark removal using Gemini 2.0 Flash have been shared on Reddit and X, where various types of watermarks are edited out and replaced seamlessly. The AI-generated images maintain a watermark in the lower left corner of the output, reflecting the emblematic nature of AI image generation. However, TechCrunch reported that this watermark removal feature is still categorized as “experimental” and is limited to tools like AI Studio.
Concerns about the unregulated capabilities of Google Gemini 2.0 Flash extend to its ability to generate images of celebrities and copyrighted characters, which it does with minimal resistance. This lack of restrictions is particularly notable considering previous incidents where a prominent artist faced undesirable image generations attributed to AI.
Featured image credit: Pawel Czerwinski/Unsplash