Google's SynthID watermark has successfully exposed its first notable deepfake: a fabricated image of Mitch McConnell.

Google's SynthID watermark has successfully exposed its first notable deepfake: a fabricated image of Mitch McConnell.

      TL;DR: Google’s SynthID watermark helped debunk a viral fabricated image of Senator McConnell in a hospital. The watermark survived screenshotting across various platforms, marking its first major successful application in the real world.

      This week, Google’s SynthID watermarking system achieved a significant milestone when Snopes used it to dismiss an AI-generated image of Senator Mitch McConnell that had gained traction on Reddit and X. The image, which depicted McConnell connected to tubes in a hospital bed and appearing to be in distress, was identified as AI-generated after Snopes discovered the embedded invisible SynthID watermark.

      Since his hospitalization following an emergency call on June 14, there has been rampant speculation regarding McConnell’s health, as he has remained mostly out of the public eye. In this instance, the claims were entirely fabricated, and the watermark functioned precisely as intended.

      SynthID, which was introduced at Google I/O in 2025, embeds an invisible signature within AI-generated images. This signature is part of the image data itself, allowing it to endure screenshots, resizing, and compression on various platforms. This durability was crucial in debunking the McConnell image, as it was shared widely, yet the watermark remained intact. OpenAI joined the SynthID initiative in May 2026, incorporating the watermark with C2PA metadata in its image creations, though Anthropic is not a participant.

      The primary limitation of SynthID is that it only functions when the image-generation tool is actively involved. Since their launch, Gemini models have included the watermark. Users can verify images by querying a Gemini model or uploading them to OpenAI’s public verification tool. However, images generated by tools outside of the initiative, such as open-source models and those from non-participating organizations, do not feature a watermark. Additionally, YouTube has initiated the auto-labeling of AI-generated videos, but the overarching issue remains that watermarking can only be effective with broad adoption, which is still voluntary.

Other articles

A £35 billion scheme to construct 14 small nuclear reactors in the UK. A £35 billion scheme to construct 14 small nuclear reactors in the UK. Michał Sołowow's SGE intends to install 14 GE Vernova Hitachi BWRX-300 small modular reactors at three locations in the UK, representing a £35 billion investment in power for the AI era, to be completed by 2034. A professor at Brown has demonstrated evidence of widespread cheating using artificial intelligence. A professor at Brown has demonstrated evidence of widespread cheating using artificial intelligence. A professor at Brown shifted his exam to an in-person format, resulting in the class average plummeting from 96 to 48, transforming a vague concern regarding AI-assisted cheating into a concrete statistic. Adult content creators targeted by scammers are having government websites taken down. Adult content creators targeted by scammers are having government websites taken down. Fraudsters are exploiting the names of adult creators to infect reputable government and university websites, as copyright claims discreetly drive thousands of harmful pages out of Google's search results. Roborock brought the excitement of the FIFA World Cup to Miami, and it was an amazing experience. Roborock brought the excitement of the FIFA World Cup to Miami, and it was an amazing experience. Roborock transformed Miami's Aventura Mall into a distinctive celebration of technology and football, creating a remarkable event. Samsung intends to establish a floating AI data center by the year 2028. Samsung intends to establish a floating AI data center by the year 2028. Samsung Heavy Industries aims to create a dedicated, 50MW floating data center by 2028, as challenges such as limited land, water availability, and local resistance hinder AI initiatives on land. Nvidia collaborates with chip competitor d-Matrix on inference technology. Nvidia collaborates with chip competitor d-Matrix on inference technology. Nvidia is integrating its GPUs with d-Matrix's inference chips in a collaborative system, marking a recent shift towards partnering with competitors in the chip industry rather than competing against them.

Google's SynthID watermark has successfully exposed its first notable deepfake: a fabricated image of Mitch McConnell.

Snopes employed Google's invisible SynthID watermark to verify the authenticity of a viral AI-generated image depicting Senator McConnell in a hospital bed. The watermark remained intact even in screenshots.