Apple’s iPhone AI Watermark Has A Catch: Google’s Gemini May Not Read It

Apple’s latest iOS 27.2 update now automatically labels some AI-edited photos with invisible watermarks to help you and others determine when an image has been edited or created with AI.

To do this, Apple is borrowing Google’s SynthID technology which already performs a similar function on the Pixel range, and applying it to its iOS-native apps.

Apple is the latest company to join partners OpenAI, Nvidia and Kakao alongside Google, all of which now embed SynthID watermarks in AI-generated or AI-edited content across “generative AI consumer products.”

Here are the ways you can use the embedded watermarks to identify AI-edited content using Google’s tools (and which may not work for iPhone images), the information Google keeps when you check an image, and a list of which iPhone apps currently use the watermarking tech.

An early report from MacRumors says that everything made in Apple’s Image Playground app carries the watermarks, while Photos embeds SynthID according to the specific functions used. The Extend and Spatial Reframing tools use it, but the Clean Up function applies it only on larger edits.

A key issue here is that many kinds of photo edits that might previously have gone unnoticed may now be spotted automatically by online services. According to The Verge, images edited with Apple’s AI-powered tools may now appear flagged as AI-edited on platforms like Instagram. It’s not clear, though, how Instagram detected the AI modification. AI images can also be labeled via metadata such as C2PA and IPTC tags, and Meta’s documentation doesn’t mention the use of SynthID for this purpose.

Which Google Tools Can Check Apple’s AI Watermark

Is It Safe To Upload Personal Photos To The SynthID Detector Or Gemini?

Google claims that photos uploaded to the SynthID detector are “processed in real-time” and deleted after the result shows. However, Google does retain a “digital signature” of your media for 24 hours to enforce daily usage quotas and defend against automated attacks.

ForbesGoogle Pulls Earth AI Tool After It Faked A Nuclear Plant In IranBy Paul MoncktonForbesCan You Tell If An Image Was Made By AI? Google Just Gave Everyone A Free ToolBy Paul Monckton