Invisible Watermark Technology: Resilient Against Gemini 2.0's Watermark Removal Capabilities

Google Gemini 2.0 Removes Visible Watermarks, But Our Invisible Protection Stands Strong

TechCrunch recently reported that users of Google's new Gemini 2.0 Flash AI model have discovered a controversial capability: removing visible watermarks from images, including those from Getty Images and other stock photo providers. While this development raises serious concerns for content creators and copyright holders, our invisible watermark technology provides a solution that remains effective even against these advanced AI tools.

The New AI Threat to Image Protection

On March 17, 2025, TechCrunch revealed how users on social media have been using Google's Gemini 2.0 Flash to not only remove visible watermarks but also to fill in gaps created by the watermark's deletion. Unlike other AI-powered tools that perform similar functions, Gemini 2.0 Flash is both exceptionally skilled at this task and free to use, making it accessible to anyone.

Although Google has stated that "using Google's generative AI tools to engage in copyright infringement is a violation of our terms of service," this policy statement does little to prevent the actual misuse of the technology. When copyright protection relies solely on visible watermarks, content owners now face a significant vulnerability.

Why Traditional Watermarks Are No Longer Enough

For years, visible watermarks have been the standard method of protecting digital images from unauthorized use. These transparent overlays typically display a copyright notice, company logo, or photographer's name. However, with AI technology advancing rapidly, these traditional protections have become increasingly vulnerable.

Before Gemini 2.0, removing watermarks required specialized skills and software. Now, anyone with access to Google's AI tools can potentially strip away these visible protections with a simple text prompt, leaving content creators exposed to unauthorized use and copyright infringement.

Imatag's Invisible Watermark: Undetectable, Unremovable

This is where our invisible watermark technology by Imatag stands apart. Unlike visible watermarks that can be targeted by AI removal tools, our technology embeds protection at the pixel level, integrating the watermark information into the image's frequency domain in ways that remain imperceptible to the human eye – and resistant to AI manipulation.

We've tested our technology against Gemini 2.0 Flash on several images with promising results: even after the AI successfully removes visible watermarks, our invisible watermarks remain intact and detectable. This test was performed using ProofPortal, which can detect our invisible watermarks even after AI manipulation. We're happy to demonstrate this capability upon request - interested parties can contact us at https://pages.imatag.com/proof-portal for a personalized demo.

ProofPortal: Verify Authenticity in the Age of AI Manipulation

ProofPortal, developed by Data Trails and utilizing Imatag's invisible watermark technology, takes this protection further by enabling content owners to verify the authenticity and origin of their visual content. It's important to note that this testing requires images to be previously registered (and therefore watermarked by Imatag) in ProofPortal. When an image with our invisible watermark is uploaded to ProofPortal, the system can:

• Confirm whether the image contains the original invisible watermark
• Detect if the image has been illegally modified
• Provide certification of the image's source and ownership
• Generate authentication reports usable in legal proceedings

This verification capability is increasingly critical in a world where AI can convincingly alter images and remove traditional protection measures.

How Our Invisible Watermark Technology Works

Our proprietary invisible watermark technology works by:

1. Analyzing the unique characteristics of each image
2. Embedding identification data using a mathematical algorithm that disperses information across the entire image
3. Creating a signature that persists through compression, cropping, color adjustments, and now – AI manipulation
4. Implementing tamper-proof protection where detection is only possible with the decryption key stored securely in the system

Unlike visible watermarks that can be targeted precisely because they're visible, our protection remains undetectable to both human eyes and AI systems attempting to remove it.

Beyond C2PA Metadata: Why Invisible Watermarks Matter

In discussing image protection, it's worth noting that industry standards like C2PA (Coalition for Content Provenance and Authenticity) – an important initiative led by companies including Adobe – recognize that metadata alone can be vulnerable. While these standards provide valuable frameworks for content authentication, they also acknowledge that metadata can be removed during image manipulation.

This is why C2PA includes support for invisible watermarks as a complementary "soft binding" mechanism. Unlike metadata that exists separately from the image itself, invisible watermarks are integrated into the pixels, creating a more persistent form of identification that survives even when an image is manipulated or its metadata is stripped.

Protecting Your Visual Content in the Gemini 2.0 Era

As AI tools like Gemini 2.0 Flash continue to advance, content protection strategies must evolve. Relying solely on visible watermarks or metadata-based solutions is no longer sufficient for professional content creators, media organizations, and stock photo agencies.

Our invisible watermark technology and the ProofPortal verification system provide the robust protection needed in this new landscape – ensuring that even as AI advances, your ownership rights remain secure.

Ready to protect your visual content against AI watermark removal? Contact us today for a demonstration of our invisible watermark technology and ProofPortal verification system.

‍