
What happened
In a hands-on test, Pangram AI Detection labeled ChatGPT images as "AI Detected" via embedded C2PA data and rated a Qwen Image 2.1 picture "High" for AI-likeness.
Why it matters
The extension correctly caught both generated images, but its maker says results are not guaranteed 100% accurate, so they should be treated as reference information only.
Summaries like this, every morning.
Pangram AI Detection offers two paths for judging an image. When a file carries the C2PA identifier, the extension can treat that metadata as a strong signal, which is how the ChatGPT image was flagged as "AI Detected." For images without that embedded marker, it falls back to analyzing the picture itself with a model called Pangram Image 1.0 and producing an "AI-likeness" score. In the test, a Qwen Image 2.1 picture without C2PA data scored "High," and the extension displayed a heat map that was almost entirely red, suggesting it judged the whole image as AI-like.
The test also covered a genuine camera photo, which was labeled "Likely NO AI" with an all-green heat map. These results suggest the tool can catch both fully generated images and edits made to real photos, though its output is not guaranteed to be 100% correct and should be used as one piece of reference information.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Puente added manga production to its Puente AI Anime Factory workflow, producing 24 pages of storyboard-style…

Brandon Thomas, founder of ArtCraft, argues that invisible tracking signals he calls "spymarks" go beyond visi…

The guide found that nano-banana-2 (RunComfy) tops installs at 440,342 and ai-video-generation follows at 385…

Ziva.sh published a walkthrough of its in-editor agent Ziva, which on Pro or Ultra plans turns a plain-languag…

A guide explains Live3D Face Swap's two easily confused modes

A Zenn article lays out a four-layer prompt template — subject, shooting conditions, typography, usage constra…
