Google Gemini 2.0 Flash sparks controversy: AI easily removes image watermarks

Google's latest artificial intelligence model Gemini2.0Flash has caused controversy because it was found to easily remove image watermarks, including professional watermarks from well-known galleries such as Getty Images. This feature concerns creators and copyright holders.

Gemini2.0Flash's image generation feature expanded access last week, but lacked the necessary restrictions. Users found that it not only generates images containing celebrities and copyrighted characters, but also intelligently fills blank areas after removing the watermark. Although there are other AI watermark removal tools on the market, Gemini2.0Flash has performed particularly well and is currently open for free, increasing copyright risks.

It is worth noting that the feature is marked "experimental" and "non-production use" and is only available in developer tools. Additionally, it does not work well when dealing with translucent or large-area watermarks. By contrast, models such as Anthropic's Claude 3.7 Sonnet and OpenAI's GPT-4o explicitly reject the removal of watermark operations, which Claude even defines as "immoral and potentially illegal". Under U.S. copyright law, removing watermarks without permission is often considered illegal.

As of now, Google has not responded to this incident. This controversy once again triggered in-depth thinking on the boundaries of AI technology and the balance of creators' rights and interests.