Google has opened to the public SynthID Detectoryour tool for identify images, videos and audios generated with artificial intelligence. The service is now available worldwide in English through synthid.comafter a first stage in which its access was limited mainly to journalists, researchers and media professionals.
Who wants to carry out a verification you just have to upload the file to be analyzed and wait for the system to look for a SynthID watermark. This signal is incorporated during the creation of the content and is imperceptible to people, but can be recognized by the detector.
The tool allows you to identify content generated with AI from Google and other companies that have adopted this technology, such as OpenAI, NVIDIA and Kakaowhile Apple It will be added soon, as announced by Google.
The operation of the detector is limited in that if a file comes from an AI that does not incorporate these marks, SynthID will not be able to identify it. Therefore, a result indicating that it is not AI does not necessarily prove that it is a real photograph or an authentic recording.
The marks are designed to resist common changes to files such as cropping, compression or the application of filters. Even so, Google recognizes that successive modifications can make it difficult to detectso it does not guarantee that the signal will be detected in any circumstance.
Google introduced SynthID in 2023 to mark AI-generated images and later extended its use to video and audio. Since then he has incorporated these signals into more than 180 billion images and videosplus an audio volume equivalent to 240,000 years of reproduction.
SynthID Detector This expands the verification options that Google already offers in other products. Gemini, the Google search engine and Chrome allow you to upload content and verify if it has been generated with AI. Together, they process more than one million verification requests per day.