University of Maryland Paper Finds Frontier Models Can Identify AI Writing and Author Identity from Text
A new University of Maryland paper finds that frontier models can detect AI-generated writing at a level comparable to specialized detectors such as Pangram, while also explaining their reasoning and identifying a text's author or the lab behind its generating model. The study provides models with only a single text and a zero-shot prompt, without web searches, tools, or candidate lists. Claude Fable 5.2 identified 56% of the reviewers behind 398 private peer reviews whose authors were not disclosed, and matched 69% of Reddit users who had posted more than 500 times in a year to their corresponding accounts based on their comments.
Why read this
SelectedAI score 80
The study applies zero-shot text source identification to anonymous peer review and pseudonymous accounts, offering a concrete approach to risk testing for products that rely on writing anonymity.
Topics
Body
New University of Maryland paper finds that frontier models now detect AI-written prose as well as trained detectors like Pangram, while also explaining their reasoning.
Can name a text's author, AI lab or specific person, from that text alone.
Anonymous peer review and pseudonymous accounts assume writing alone doesn't reveal who you are. Spotting AI text was also thought to need trained detectors, not plain prompting.
The authors gave models a single text and a zero-shot prompt, with no web search, tools, or candidate list. They tested AI-or-human detection, which lab's model wrote it, and which person wrote it.
Claude Fable 5.2 named the reviewer of 56% of 398 private peer reviews whose authors were never made public. It named the account behind 69% of comments from Redditors with over 500 posts in a year.
If your product depends on anonymous writing, such as blind review or pseudonymous forums, treat author anonymity as weak and test it against frontier models.
AI Trends