Google has launched the public version of its SynthID Detector website, enabling users to verify if image, video, or audio files contain imperceptible watermarks embedded by Gemini and partner models like OpenAI, NVIDIA, and Kakao.
The platform requires a Google, OpenAI, or Apple account sign-in and enforces a limit of roughly ten checks per day to prevent bypass attempts, while remaining unable to scan models lacking native SynthID support.
The tool is called SynthID Detector and it works in a very simple way. When someone uploads a file to the dedicated website, the system scans the content for a special, invisible mark left by artificial intelligence tools. To use it, someone must log in with a Google, OpenAI or Apple account.
This technology first appeared in 2023, while in 2025 a limited version was released exclusively for researchers and journalists. Now, the tool is open to the general public.
In addition to Google's system, the tool can also detect signs from partner companies, including OpenAI, NVIDIA, and Kakao. Apple is also participating in the program, although support for recognizing its own creations is still under development.
READ ALSO
How Apple's Image Playground works to create images with artificial intelligence
The limitations of the tool
Despite its convenience, the system has significant limitations. Because it only looks for specific hidden marks, it cannot detect material created by third-party systems, such as Meta's tools, which do not use the specific markup. The tool also only shows whether the mark is present, without revealing which parts of the file were created by artificial intelligence.
According to Google, these hidden marks remain even if the file is edited after it is created. To prevent attempts to bypass or remove the marking, the company has set a limit of around ten checks per day per user.
G
Hacks.gr on Google
Set it as a preferred source for cybersecurity updates.
Preferred source
Top comments (0)