AI tracked down nearly 80,000 ‘ghost students’ trying to enroll in California colleges

Ghost students are fake or stolen identities practiced by fraudsters who overwhelm colleges with thousands of requests in minutes, and they earn quickly as students, request financial assistance, and then disappear with money. Some even offer homework to keep the preamble from separation before they can collect. The Ministry of Education has issued a consultant in the spring to be colleges at a state of maximum alert for fraudulent students, who not only steal financial aid, but also occupy a large space in the classroom so that real students cannot enter into the training courses they need to graduate. Last month, the Ministry of Energy revealed that it had found 150,000 identity suspected of federal aid models, and said 90 million dollars for unqualified students. It follows $ 30 million of aid provided to the dead whose identities were used to register in the classroom.

After being firmly exposed in 2024 by Ghost students, California Community College (CCC) began to fight the artificial intelligence scheme-with artificial intelligence. This month, CCC launched an Amnesty International initiative at the Foundation level, using the N2N Lightlet.AI platform to detect fraudulent records. Since the operation, which is still in effect in all 116 colleges, 79,016 total applications were discovered as fraudulent across more than half a million requests, according to the last update.

“The only answer to a bad man with artificial intelligence is a good man with artificial intelligence.” luck.

Kodithala is the CEO and founder of N2N Services, and the LightLeap.AI platform is now part of the CCC registration system to detect fraud and registration in class and financial aid operations. The system will be deployed through the full CCC system by the end of the 2025 calendar year.

So far, the ghost student epidemic diverges in the regions, but even the largest is not immune to attacks.

The Santa Barbara Community College area witnessed 320,487 applications that were treated and 24,485 students who were marked at the price of a fake student by 7.6 %, while Vintera province witnessed a fake student rate of 21.4 % through 113,204 requests that have been treated. The citrus area had a rate of 34.6 % across 49,837 requests that were treated. In the smaller lassen area, 65.3 % of the requests – 4,652 out of 7,129 – were marked by Lightleap as fraudulent.

Kodithala said that the colleges of society are specifically goals because they are open educational and educational institutions that are designed to accept any person to a large extent.

He said: “Community colleges are open access – just like physical universities where anyone can walk in it. This is according to the design.”

Kodithala added that traditional identity methods do not work well with a population group consisting of 18-24 years. This age group often does not have a first credit card, company function, email address, or property records. It creates a loophole that is not exploited by fraudsters who use ghost students. Kodithala said that the LightLeap platform, therefore, is looking at databases to verify identities using these entry points, but it is also looking into the application data to know the patterns of what the model fraud students do when they try to register in the community college.

“They are following a less resistance path,” said Kodithala. “They fill enough fields of application to bypass screens and not fill more than required. The ordinary student lists all their previous addresses and all their previous schools.”

Outside California, Lightletap works in schools in Arizona, Michigan and Minnesota. However, the main pain point is to ensure that a sign of real students is not somewhat put as fraudulent, which makes it difficult for them to apply and register in the college. On the contrary, fraudulent students are marked as an uncontrolled, and they occupy a class and resources in class and goal.

“Think about the subway sandwich store,” said Kodithala. “Thousands of people put a sandwich and 3000 real people, but 700 bad requests. You lose money on bad requests and less able to serve good people.”