Google Unveils SynthID.com to Detect Invisible AI Watermarks in Images and Videos

Introducing SynthID.com

According to НВ — Техно: Google has launched SynthID.com, a platform designed to verify images and videos for an invisible AI-generated watermark. This service currently supports content from a select group of partners, including well-known names like Google, OpenAI, Nvidia, and Kakao. Users can upload photos or videos to check for the SynthID watermark-a hidden marker that remains intact even after editing or cropping, unlike traditional visible watermarks.

How SynthID Technology Works

Developed by Google DeepMind in 2023, SynthID identifies media created with tools from Google and its partners who implement this technology. In one test, the system successfully flagged a fake AI-generated image featuring Donald Trump, Ken Paxton, and McDonald's Grimace, created using the Gemini AI. Although SynthID detected the image as artificially generated or altered, Gemini mistakenly claimed the fake was based on a genuine photograph of Trump and Paxton from March 2024-a photo journalists could not verify.

Another experiment with an image produced by Grok, a different AI tool, showed SynthID did not detect its watermark and assessed a low likelihood that the content was created using Google partner AI. The system warned that the image might have been generated by another company or produced before SynthID's adoption. Notably, SynthID technology is already integrated across various companies' model ecosystems, with its list of partners steadily growing.

Access to SynthID.com requires users to sign in using a Google, Apple, or ChatGPT account.

The launch of SynthID.com highlights the increasing need to authenticate digital content as AI-generated images and videos become more widespread.

This watermarking solution offers valuable support for journalists, researchers, and professionals dedicated to verifying the authenticity of digital media. The expanding network of SynthID partners reflects the rising demand for effective tools to combat misinformation in the digital age.

The need for advanced verification methods is increasingly evident, as illustrated by the recent developments in AI watermarking technology. For instance, Google DeepMind has also introduced innovative solutions like invisible watermarks in synthetic proteins, showcasing the expanding applications of watermarking beyond traditional media. This highlights the broader implications of authentication in the digital landscape.