Users on social media have discovered a disputed employ case for Google’s new Gemini AI model: removing waterlabels from images, including from images rehireed by Getty Images and other well-understandn stock media outfits.
Last week, Google enbiged access to its Gemini 2.0 Flash model’s image generation feature, which lets the model natively originate and edit image satisfyed. It’s a mighty capability, by all accounts. But it also ecombines to have scant protectrails. Gemini 2.0 Flash will unprotestingly originate images depicting celebrities and imitaterighted characters, and — as alluded to earlier — delete waterlabels from existing pboilingos.
New send unlocked: Gemini 2 Flash model is reassociate awesome at removing waterlabels in images! pic.twitter.com/6QIk0FlfCv
— Deedy (@deedydas) March 15, 2025
As disjoinal X and Reddit employrs remarkd, Gemini 2.0 Flash won’t fair delete waterlabels, but will also try to fill in any gaps originated by a waterlabel’s deletion. Other AI-powered tools do this, too, but Gemini 2.0 Flash seems to be exceptionassociate sended at it — and free to employ.
Gemini 2.0 Flash, useable in Google’s AI studio, is amazing at editing images with plain text prompts.
It also can delete waterlabels from images (and puts its own reserved waterlabel in instead 🤣) pic.twitter.com/ZnHTQJsT1Z
— Tanay Jaipuria (@tanayj) March 16, 2025
To be clear, Gemini 2.0 Flash’s image generation feature is taged as “experimental” and “not for production employ” at the moment, and is only useable in Google’s prolonger-facing tools enjoy AI Studio. The model also isn’t a perfect waterlabel deleter. Gemini 2.0 Flash ecombines to struggle with certain semi-see-thharsh waterlabels and waterlabels that canvas big portions of images.
Still, some imitateright helderlyers will certainly apshow rehire with Gemini 2.0 Flash’s informage of usage remercilessions. Some models, including Anthropic’s Claude 3.7 Sonnet and OpenAI’s GPT-4o, clpunctual decline to delete waterlabels; Claude calls removing a waterlabel from an image “unmoral and potentiassociate illterrible.”
Removing a waterlabel without the distinctive owner’s consent is pondered illterrible under U.S. imitateright law (according to law firms enjoy this one) outside of unwidespread exceptions.
Google didn’t promptly react to a seek for comment sent outside of standard business hours.
Updated 3/17 at 1:48 p.m. Pacific: A Google spokesperson provided the follotriumphg statement:
“Using Google’s generative AI tools to join in imitateright infringement is a violation of our terms of service. As with all experimental liberates, we’re seeing shutly and hearing for prolonger feedback.”
