Google’s SynthID Detector Recognizes Files from OpenAI, NVIDIA, and Kakao

Google has updated its SynthID Detector tool, expanding its coverage. The service no longer checks only files created with the company’s own artificial intelligence (AI) models. It also recognizes image, video, and audio files produced by models from OpenAI, NVIDIA, and Kakao. AI content created with Apple will reportedly be added to the system soon.

SynthID adds an invisible digital marker to content generated or modified with artificial intelligence. SynthID Detector finds this marker and indicates that the file was created using AI. The system, launched by Google in 2023, has watermarked more than 180 billion images and videos to date. In addition, audio content with a total length reaching 240,000 has also been marked.

To use the tool, users must visit synthid.com. They can sign in to the site with a Google, Apple, or ChatGPT account. The image, video, or audio file to be checked is then uploaded to the system. SynthID Detector indicates whether it has found a matching watermark in the file. If the marker is detected, the content is identified as AI-generated.

SynthID Detector is not a system capable of recognizing all AI-generated content. The tool only searches for a SynthID watermark previously added to the file by supported models. If an AI-generated file does not contain this marker, its origin cannot be determined. Therefore, a negative result does not confirm that the content was created by a human.