AI in Education AI Minute Newsroom 2026-08-10

The professor noticed his history class had filled with students who did not exist — and were doing the homework

At East Los Angeles College, the history professor David Song began seeing names on his rosters that did not fit the campus: generic Anglo-Saxon names in a student body that is mostly Latino and Asian, with academic backgrounds that made no sense. They were not students. They were 'ghost students' — enrolments created by fraud rings using stolen identities, filed in bulk by bots, kept alive by AI submitting just enough coursework to survive the census date, and cashed out as state and federal financial aid. The New Yorker returned to the story on 9 August. The numbers behind it are large and public: about a third of applications to California's 116 community colleges in 2024 were fraudulent, the state has lost more than $30 million since then, and federal investigators have looked into more than $350 million in ghost-student fraud nationally over five years. The counter-attack is also AI. California's community college system now runs applications through automated fraud detection — one platform, N2N's LightLeap.AI, has flagged more than 79,000 applications — and colleges report the software catching roughly twice as many fraudulent enrolments as human staff, with some campuses estimating detection above 90 per cent. Monthly losses have fallen from the 2025 peak to around half a million dollars. Individual campuses have added live human verification for enrolment.

Why it mattersEvery debate about AI in education is about students cheating on assignments. This is the same technology aimed one level up — not at the coursework but at the enrolment system itself, and not by students at all. The victims are threefold and none of them is the college: real people whose stolen identities now carry loans they never took, real students who cannot get a seat in a full class where most of the seats are held by nobody, and the public purse. What makes this the sharpest case study available is that both sides are automated. Bots file the applications; classifiers screen them; a fraud ring adjusts; the detector retrains. The humans left in the loop are professors like Song, who noticed something the system did not because he knew what his own classroom looks like. That is the durable lesson for anyone running an institution with an online front door: the automated defence is now genuinely better than the manual one at volume, and it still needed a person on the ground to know that something was wrong in the first place.

✓ Verified · 4 sources

The professor noticed his history class had filled with students who did not exist — and were doing the homework

Related stories