Google’s Gemini AI model, particularly Gemini 2.0 Flash, has drawn attention for its ability to remove watermarks from images, including those from notable stock websites like Getty Images. While this experimental feature is designed for non-commercial use, its effectiveness in watermark deletion raises concerns among copyright holders who might pursue legal action. Additionally, the model’s text-to-image capabilities lack significant restrictions, allowing it to generate images of celebrities and other copyrighted content. As Google promotes transparency through initiatives like SynthID for watermarking AI-generated text, challenges remain regarding adoption and the effectiveness of content provenance standards developed by the C2PA coalition.
[ad_1]
Reports suggest that Google’s Gemini AI model is being utilized to eliminate watermarks from images, despite the company’s efforts to clarify the labeling of content generated by AI.
Users on platforms like X and Reddit have noted that Gemini 2.0 Flash has the ability to remove watermarks from images sourced from Getty Images and other prominent stock media sites. According to a report by TechCrunch, this experimental and free-to-use image generation functionality of Gemini 2.0 Flash is more proficient at deleting watermarks compared to other AI tools.
In addition to its watermark removal capabilities, the model’s text-to-image generation feature seems to have fewer restrictions, allowing it to potentially create images of celebrities and other copyrighted material.
The supposed absence of usage restrictions and the capability to remove watermarks without permission might lead to increased legal challenges from copyright owners.
It’s important to note that access to Gemini 2.0 Flash’s image generation feature was broadened last week for “experimental” use and is “not for production.” Currently, it is only available through Google AI Studio, the company’s platform for AI developer resources.
Earlier this year, Google DeepMind revealed its plan to make SynthID Text, a tool for watermarking AI-generated text, publicly accessible. Last month, it also announced that its Google Photos app would begin incorporating SynthID to label AI-edited images.
The Alphabet-owned firm is also a participant in the Coalition for Content Provenance and Authenticity (C2PA), alongside companies like Amazon, Microsoft, OpenAI, and Adobe.
Story continues below this ad
The C2PA has established technical standards designed to trace the origins of AI-generated images, offering metadata that includes information such as the creation date and the AI tools used to create the content. However, it faces challenges in terms of widespread adoption and interoperability.
© IE Online Media Services Pvt Ltd