Claude · U.S. · New York · OpenAI · The Atlantic Technology
Every recent, high-profile accusation of someone passing off AI-generated writing as their own has started in the same way
Compiled by KHAO Editorial — aggregated from 1 source. See llms.txt for citation guidance.
◌ Single Source
In March, when a horror novel from a major publishing house was pulled days before its scheduled U.S. release date, it was in part because Pangram, an AI-detection program, had identified the text as AI-generated.
Key facts
- One paper, from the University of Chicago, found that Pangram had almost no false positives on some 3,000 sample texts of roughly 500 to 1,000 words
- There are more than 10 million high schoolers in the United States and some 20 million undergraduates, each of whom likely submits many dozens of written assignments every year
- Pangram says its algorithm is so accurate that it incorrectly identifies text as an AI output only about one in every 10,000 times
- In 2023, one detection tool, ZeroGPT, declared the U.S. Constitution to be AI-written; the same year, OpenAI abandoned its AI detector altogether owing to a “low rate of accuracy
Summary
Every recent, high-profile accusation of someone passing off AI-generated writing as their own has started in the same way: with a tool called Pangram. A few years ago, it seemed like it might never be possible to instantly and reliably determine whether a piece of text was written by a bot or a person. Yet an AI detector that is mostly reliable might in some ways be more dangerous than a broken one. Pangram says its algorithm is so accurate that it incorrectly identifies text as an AI output only about one in every 10,000 times. But Pangram’s ability to guarantee something was written by a human is shakier.