Google Gemini 2.0 Flash's Watermark Removal Feature Draws Attention

Recently, social media users discovered that Google's new AI model can remove watermarks from images, including those from established names like Getty Images and other leading stock photo libraries. This development has raised the dust of controversy, quickly gathering attention from all quarters. 

Powerful, yet Controversial 

Last week, Google expanded access to the image generation capabilities of the Gemini 2.0 Flash model, enabling the model to natively generate and edit image content. This is undoubtedly a powerful feature, but some users have observed certain limitations. Gemini 2.0 Flash can seamlessly generate images of celebrities and copyrighted characters, and remove watermarks from existing photos.

Users on X (formerly Twitter) and Reddit pointed out that Gemini 2.0 Flash not only removes watermarks but also attempts to fill in the resulting blank spaces. Other AI-powered tools also have this function, but Gemini 2.0 Flash appears to be particularly specialized at it, and it's free.

Still in the Works 

Users must note that the image generation feature of Gemini 2.0 Flash is labeled "experimental" and "not suitable for production," and is only available in Google developer tools like AI Studio. The 2.0 Flash model isn't a perfect watermark removal tool either, as it seems to struggle with some semi-transparent watermarks and those that cover a large portion of the image.

Ethics and Copyright Issues 

This development has made copyright holders question the lack of usage restrictions in Gemini 2.0 Flash. Models like Anthropic's Claude 3.7 Sonnet and OpenAI's GPT-4 explicitly refuse watermark removal, with Claude stating that removing watermarks without the image owner's consent is "unethical and potentially illegal." Under US copyright law, removing watermarks without consent is illegal in most cases.

Google is yet to respond to comment on these issues.