Valar Intelligence unveiled Puddin AI in Osaka on June 14, marketing a detector that flags synthetic academic prose by analyzing how quickly document revisions appear.
Engineers said the tool inspects word-processor metadata rather than attempting unreliable fingerprinting of model-generated sentences alone.
Japanese universities piloting the software combine Puddin AI with traditional plagiarism checks before referring cases to disciplinary committees.
Critics warned that rapid editing tools could trigger false positives for students who dictate notes or collaborate on shared documents without cheating.
Academic publishers said updated submission policies would cite layered detection approaches similar to Puddin AI when reviewing manuscript authenticity.
Faculty senates in Japan are piloting Puddin AI alongside plagiarism tools, seeking layered evidence before accusing students of unauthorized model use.
Valar engineers said the product analyzes metadata trails left by word processors rather than attempting to fingerprint model outputs directly.
International education groups cautioned that detection arms races could punish students who legitimately use approved editing assistants without clear institutional policies.
Publishers of academic journals said they would issue updated author guidelines referencing AI detection tools like Puddin AI.
Created by Ayen Stabel.
Stabel is AI and can make mistakes.
Sources:
https://english.kyodonews.net/articles/-/77851