Flawed AI detectors like Turnitin trigger legal battles
As educators and publishers increasingly rely on unreliable AI detection tools like Turnitin and GPTZero, a wave of false accusations is sparking lawsuits and forcing schools to abandon them.
The rapid adoption of generative AI has triggered a parallel boom in AI detection software, but these tools are now facing severe backlash. A Center for Democracy and Technology survey found that 43 percent of US teachers in grades six through 12 regularly used AI detectors between 2024 and 2025. Services like Turnitin, which automatically enabled AI detection in 2023, claim false positive rates of under 1 percent. Pangram claims a rate of 1 in 10,000, while GPTZero also boasts low error rates. Yet, these claims are increasingly challenged.
The consequences of relying on these algorithms are escalating. In February, an Adelphi University student won a lawsuit against his school over AI allegations. Similarly, Thierry Rignol sued Yale University after a professor used GPTZero to scan his exam, leading to a suspension. In publishing, Minotaur canceled a $2 million book deal with author Jerry Falade over AI concerns. Public figures are also weaponizing these tools; Jack Osbourne recently used a detector called Getsolved to publicly accuse journalist Kat Tenbarge of using AI.
These incidents highlight deep technical flaws. A 2023 Stanford study revealed that detectors disproportionately flag writing by non-native English speakers. Tools like QuillBot look for predictable sentence structures, which often overlap with the natural writing styles of non-native or neurodivergent writers. The high rate of errors even led OpenAI to shut down its own detector in 2023. Consequently, institutions like Yale, Johns Hopkins, Vanderbilt, and Georgetown have restricted these tools, while the Massachusetts Institute of Technology warned they do not work.
For developers and educators, this shift highlights the futility of relying on algorithmic policing. Instead of automated detectors, organizations are shifting toward alternative verification. Platforms like Substack have integrated Pangram, while LinkedIn added a "seems like AI slop" button. However, the industry is moving toward human-centric solutions. The Authors Guild is now issuing "Human Authored" certifications to help writers protect their reputations, signaling that the future of content verification lies in proactive certification rather than flawed automated detection.
This is our own summary of reporting by The Verge AI
