Pangram Has Emerged as the Gold Standard of AI Detection. Should You Trust It?
Meet the AI police who can make or break careers—in publishing and beyond.
Pangram has become the leading AI detection tool, promising to determine the extent of AI involvement in generating text using a percentage-based assessment. In January, speculation arose on social media platforms that author Mia Ballard may have utilized AI to write her self-published novel, Shy Girl, which was later revealed to contain 78% AI-generated content.
This led Hachette to cancel the novel's release. Similar accusations followed, with The New York Times facing scrutiny for publishing an AI-generated installment of its Modern Love column, receiving a 100% score on Pangram. Other works, such as a Commonwealth Short Story Prize winner and a thriller sold for $2.4 million, also garnered high AI detection scores.
Despite this, not all authors endorse the use of AI detection software. Jane Friedman, an author and publishing expert, expresses frustration towards these tools, feeling they are just as harmful as the AI companies they target. Outside of publishing and academia, Pangram is not widely recognized, but its relevance is growing as the demand for distinguishing AI-generated content from human-written text increases.
Spero, one of Pangram's founders, discusses his background in software development and his collaboration with co-founder Bradley Emi before founding Checkfor.ai, which was later renamed Pangram. The company's "synthetic mirroring" technique involves LLMs generating a close match of human writing to train its model. Additionally, Pangram utilizes "hard negative mining" to search for false positives and incorporate them into its training set.
The company has gained traction in various industries, with creative writing constituting the largest segment of its training text.
The Shy Girl story brought Pangram into the spotlight, but Spero downplays its influence on book deal decisions. The AI detection tool is integrated into Substack, enabling readers to quickly gauge the likelihood of AI involvement in a piece of content.
Written by urgent.news from Wired's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
