Tools
Google lets anyone check whether something is AI, but only for content with a SynthID mark
Since October 7, anyone can upload an image, a video or an audio file and see whether it carries the invisible SynthID mark from Google or its partners. For now, in English only.
The essentials
2 confirmed facts · 2 according to the source · 3 open questions
- Google opened its SynthID detector to everyone on October 7. It is available worldwide, but for now only in English.1
- The detector recognizes images, videos and audio made with AI from Google or from partners such as OpenAI, NVIDIA and Kakao.12
- According to the sourceGoogle says it has marked more than 180 billion images and videos and 240,000 years of audio since 2023.1
- According to the sourceAccording to Ars Technica, outside Google's tester program, checking for the mark meant asking Gemini, and it only detected Google's version.3
Why it matters
Checking whether something carries an AI mark goes from a cumbersome process to a website anyone can use. This suggests a clear limit: the tool only detects content that carries SynthID. If it doesn't find the mark, that doesn't prove something is real, because it may come from an AI that doesn't use this system.
If you receive a doubtful image, video or audio file, you can upload it and see whether it carries the mark. But the website only works in English, and if it doesn't find the mark, you can't assume the content is authentic.
The details
SynthID is an invisible signal that Google's tools add to what they generate: it goes in the pixels of images and videos and in the waveform of audio, according to Ars Technica3. Until now, anyone not in Google's tester program had to ask Gemini. In addition, the earlier detector only recognized Google's version: ChatGPT images carry SynthID, but they did not show up as marked3.
The website accepts common formats, such as JPG, PNG, MP4 or MP3, according to TechCrunch2. OpenAI also maintains its own checking page. Microsoft and Meta have their own marking systems, which according to TechCrunch are not foolproof and often fail to recognize even the content of their own models2.
What we don't know
- When it will be available in Spanish or other languages: Google only mentions English for now.
- When Apple will join, which Google says will happen "soon".
- How well the detector works according to independent tests: for now, there is only data from Google.
Background
Google launched SynthID in 2023. Last year it presented an early version of the detector for media professionals, which according to TechCrunch was tested with journalists and researchers at Google I/O12.
Sources
- 1We're making it easier to identify AI-generated content globally.
- 2Google’s new SynthID website can identify AI-generated media
- 3Google rolls out improved SynthID AI content detector, now available globally
Was this story useful? Yes Not really