The ghost students are fake or stolen identities wielded by scammers who flood colleges with thousands of applications in minutes, gain fast acceptance as students, request financial aid, and then disappear with the money. Some even submit homework to keep from being booted from class before they can collect. The Department of Education issued an advisory in the spring for colleges to be on high alert for fraudulent students, who not only steal financial aid but also take up so much space in classes that real students can’t get into the courses they need to graduate. The DOE revealed last month it found 150,000 suspect identities in federal student aid forms, and said $90 million had gone out to ineligible students. It traced $30 million in aid given to dead people whose identities were used to enroll in classes.

After being hit hard in 2024 by ghost students, the California Community College (CCC) system started fighting the AI-driven scheme—with AI. This month, the CCC launched an enterprise-wide AI initiative, using N2N’s LightLeap.AI platform to detect fraudulent enrollments. Since the rollout, which is still in the process of taking effect across all 116 colleges, 79,016 total applications have been detected as fraudulent across more than half a million applications, according to a recent update. 

“The only answer for a bad guy with AI is a good guy with AI,” Kiran Kodithala told Fortune. 

Kodithala is CEO and founder of N2N Services, and his LightLeap.AI platform is now part of the CCC enrollment system to detect fraud across applications, class registration, and financial-aid processes. The system will be deployed across the full CCC system by the end of the 2025 calendar year. 

So far, the scope of the ghost student epidemic has varied across districts, but even the largest aren’t immune to the attacks. 

The Santa Barbara Community College District saw 320,487 applications processed and 24,485 ghost students flagged for a 7.6% fake student rate, while Ventura County saw a 21.4% fake student rate across 113,204 applications processed. The Citrus district had a 34.6% rate across 49,837 applications processed. At the smaller Lassen district, 65.3% of applications—4,652 out of 7,129—were flagged by LightLeap as fraudulent. 

Community colleges are targets specifically because they are open-access educational and learning institutions that are designed to accept pretty much anyone, said Kodithala. 

“Community colleges are open access—just like physical campuses where anybody can just walk in. And that’s by design,” he said. 

Traditional identity-verification methods don’t work as well for a demographic cohort composed of ages 18-24, Kodithala added. That age group often doesn’t have a first credit card, corporate job and email address, or property records. It creates a vulnerability that the fraudsters wielding ghost students are exploiting. Kodithala said the Lightleap platform, therefore, looks at databases to verify identities using those entry points, but also looks at the application data to see patterns of what typical fraudulent students do when they try to enroll in community college. 

“They follow the path of least resistance,” Kodithala said. “They fill in just enough application fields to get past the screens and not fill in any more than is needed. A regular student lists all their previous addresses and all their previous schools.”

Beyond California, LightLeap is in action at schools in Arizona, Michigan, and Minnesota. A key pain point, however, is in making sure little to no real students are flagged as fraudulent, he said, making it more difficult for them to apply and enroll in college. Or, conversely, fraudulent students are flagged as non-fraudulent, and they take up space in class and waste time and resources. 

“Think of a Subway sandwich store,” Kodithala said. “Thousands of people are putting in sandwich orders and 3,000 are from real people but 700 are bad orders. You’re losing money on the bad orders and less able to serve the good people.”