As schools crack down on student AI use, Johns Hopkins University professor Steve Hanke and other educators report they can easily detect chatbot-generated essays because the output stands out against students' typical writing standards. While commercial AI detectors suffer from false-negative rates as high as 99.6% and tech firms sell both detection and evasion tools, schools are forced to improvise. Responses range from Tasmania blocking AI networks and New South Wales proposing bans on unsupervised take-home assignments, to South Australia embracing custom classroom AI tools.
AI cheating detection methods
- ▪Some educators are responding to AI cheating by designing AI-resistant assignments or reverting to handwritten, oral, or pen-and-paper exams.
- ▪Johns Hopkins University applied economics professor Steve Hanke stated he can easily detect AI-generated student submissions because chatbot output stands out against the poor writing skills of modern students.
- ▪Former Harvard and Columbia admissions dean Drusilla Blackman stated that educators can easily detect student AI use because the submitted work fails to match the student's typical standard of writing or critical thinking.
State education policy responses
- ▪The Tasmanian Department for Education, Children and Young People blocked AI applications and websites for all students on department-owned devices and networks.
- ▪South Australia partnered with Microsoft to roll out EdChat, a custom, secure ChatGPT-style generative AI tool designed to challenge public high school students through guided questioning.
- ▪The New South Wales government proposed banning unsupervised take-home assignments for year 11 and 12 students to combat AI cheating, following over 1,000 recorded cheating cases in 2025.
AI detection tool limitations
- ▪University of Florida researchers tested five popular AI text detectors and found false-negative rates as high as 99.6 percent, with simple vocabulary tweaks successfully defeating most tools.
- ▪AI text detectors throw false positives that disproportionately flag non-native English speakers, according to research reported by Digital Trends.
Student AI dependency risks
- ▪Research indicates that student reliance on AI tools creates a vicious cycle of dependency, preventing the conversion of information into long-term knowledge.
- ▪University of Technology Sydney professor Leslie Loble warned that AI disrupts the learning process by keeping students in surface learning rather than converting working memory into deep, retained knowledge.
AI tool market contradictions
- ▪Some technology companies sell both AI detection software and the evasion tools designed to bypass them, such as Superhuman's Grammarly and GPTZero.
- ▪Superhuman education director Jenny Maxwell stated that the technological arms race between AI detection and evasion is ultimately a dead end.
Story comments
Loading comments…