Google has made its SynthID Detector available to the public, allowing users worldwide to verify whether images, videos, and audio were generated or altered using artificial intelligence (AI). This move aims to address concerns about misinformation, fraud, and credibility of online information. The detector uses Google's invisible digital watermarking technology, SynthID, to identify AI-generated content. Previously, the tool was only available to a limited group of journalists, media professionals, and researchers.
The expansion of SynthID Detector comes as sophisticated generative AI tools make it increasingly difficult for users to distinguish authentic media from synthetic content. Google DeepMind's vice president of Science and Strategic Initiatives, Pushmeet Kohli, emphasized the importance of clear context about online content to make informed decisions about what to trust. Since its introduction in 2023, SynthID has been used to watermark over 180 billion images and videos, as well as audio equivalent to about 240,000 years.
SynthID embeds an imperceptible signal into AI-generated media that can remain detectable after common modifications. Unlike conventional labels that can be removed or obscured, SynthID's watermark is designed to remain detectable even after cropping, filtering, compression, or changes to video frame rates. The technology has been integrated into Google's generative AI ecosystem, including Gemini and its image, video, and audio generation models.
The latest expansion of SynthID Detector broadens the system beyond Google-generated content, allowing it to identify watermarks associated with AI-generated content from partners including OpenAI, NVIDIA, and Kakao. Support for Apple is expected to be added in the future. This development is significant, as the proliferation of generative AI has created a fragmented environment with different AI companies using different approaches to identifying synthetic media.
Google has also integrated SynthID verification into products such as Search, Gemini, and Chrome. These built-in verification features now handle over one million requests a day. The public release of SynthID Detector reflects a wider industry push to provide more information about how digital content was created. AI-generated images, video, and audio can now be produced with a level of realism that makes visual inspection increasingly unreliable.
The development of SynthID Detector could be particularly relevant for newsrooms and businesses dealing with an increasing volume of AI-generated media. Establishing the provenance of an image, video, or audio recording can have financial and reputational consequences. As generative AI becomes more embedded in everyday digital communication, the challenge shifts from creating convincing synthetic media to establishing whether the content people encounter can be trusted.
Google acknowledges that watermarking is not a complete solution to AI-generated misinformation. SynthID can identify content carrying its watermark, but it cannot automatically establish that every piece of media without a watermark is authentic. The company has also backed the Content Credentials framework developed by the Coalition for Content Provenance and Authenticity (C2PA), which records information about how digital media was created and modified.
Key points
- Google's SynthID Detector is now available globally to identify AI-generated content.
- The detector uses invisible digital watermarking technology to identify AI-generated media.
- SynthID Detector can identify watermarks associated with AI-generated content from multiple partners.
