Apple adopts Google technology to watermark AI-edited photos

Apple integrates SynthID

Apple has introduced a new way to identify AI-generated and edited photos within its iOS 27.2 update. The company is using a technology called SynthID, which was originally developed by Google. This tool embeds invisible watermarks into images to help users distinguish between authentic photos and those modified by artificial intelligence.

By adopting this technology, Apple joins a group of major partners including OpenAI, Nvidia, and Kakao. These companies are all working to embed SynthID watermarks into various consumer products that use generative AI. The goal is to create a standard for identifying synthetic content across the digital landscape.

The watermarking is not universal across all iPhone features. Instead, it is applied specifically to iOS-native apps and certain editing functions. This move represents a shift toward automated transparency for users who interact with Apple's suite of creative tools.

How the watermarking works

The SynthID technology functions by embedding a digital mark into the media during the creation or editing process. It is not a visible stamp on the image but rather a signal that can be read by specific software. This allows the watermark to remain hidden from the human eye while remaining detectable by compatible analysis tools.

On the iPhone, the watermark is applied to specific functions within the Photos app and other native tools. For example, everything created in the Image Playground app carries these marks. In the Photos app, the watermarks are applied to the Extend and Spatial Reframing tools, as well as larger edits made using the Clean Up function.

It is important to note that SynthID does not actually

Detection on social platforms

The implementation of these watermarks means that many types of photo edits that might have previously gone unnoticed could now be automatically flagged. Online services may be able to spot these modifications more consistently as the technology becomes more widespread across different apps.

For instance, images edited with Apple's AI-powered tools may now appear as flagged content on platforms like Instagram. However, it is currently not clear how Instagram is actually detecting these specific AI modifications. Other labeling methods exist, such as C2PA and IPTC metadata tags, but Meta's documentation does not mention using SynthID for this purpose.

Privacy and data retention

When you use the official SynthID detector website, Google claims that photos are processed in real-time and deleted once the result is shown. Despite this, the company does retain a digital signature of your media for 24 hours. This signature is used to enforce daily usage quotas and to protect the system against automated attacks.

Users who sign in through a third party, such as Apple, will have their basic account identifiers kept by Google. These identifiers include the user's email address, user ID, and sign-in provider details. Google has not specified a retention period for this personal information.

Limitations of the technology

A critical limitation of SynthID is that it does not actually detect AI-generated or AI-edited media by analyzing the pixels. Instead, it simply reads the watermarks that were deliberately embedded during the generation or editing process. If a tool does not include a watermark, SynthID will not be able to identify the content as synthetic.

This means that while images from tools like Gemini or ChatGPT should be detected, other tools like ComfyUI can generate imagery without watermarks. Because these images lack the embedded signal, they will not be spotted by the SynthID detector. This creates a gap where some AI content remains completely invisible to the detection system.

Usage limits and privacy risks

While Google does not specify an exact number of checks for the SynthID detector, the Gemini help page indicates a limit of about 10 image checks, 10 video checks, and 10 audio checks per rolling 24 hours when using the app. Users should be mindful of these limits when trying to verify multiple pieces of media in a short period.

There are significant privacy considerations when using these tools for sensitive or potentially offensive media. It is advisable to avoid uploading such content online and instead seek other verification methods. This caution is especially important in the Gemini app, where images are treated as chat media that may be retained or subject to human evaluation based on your specific settings.