California Colleges Are Using AI to Catch Fake Students

California Colleges Are Using AI to Catch Fake Students

Most of the conversation about AI in admissions is about applicants using AI—to write essays, polish personal statements, or pad their record. We've tracked which colleges use AI to read and score essays, what universities spend on AI detection, and the paradox of colleges that ban student AI while using AI themselves. This is the flip side: in California, the fastest-growing use of AI in admissions is colleges turning it against a very different problem—applicants who don't exist at all.

California's community college system—116 colleges, the largest in the country—has been flooded with "ghost students": bot-generated and identity-theft applications whose only goal is to enroll, collect financial aid, and disappear. The system's answer has been to fight AI with AI.

The Problem: Ghost Students and Stolen Aid

The scale is hard to overstate. According to Chancellor's Office figures reported by CalMatters and EdSource, the share of community-college applications flagged as fraudulent climbed from roughly 20% in 2021 to about 25% in early 2024 and 34% by spring 2025—with some campuses far higher. The Los Rios district near Sacramento saw fraudulent applications hit 64% in early 2025 before AI screening pushed it back down toward 12% a year later.

The money follows the fake applications. By the Chancellor's Office's own accounting:

• More than $30 million in financial aid has been stolen since 2024 (an earlier estimate put losses since 2021 at $18M+, and the office called that an undercount).
• 2024 alone: $11M+.
• The peak stretch, January–May 2025, saw roughly $13 million disappear.
• Since AI screening rolled out in summer 2025, the run rate has fallen to about $500,000 per month.

These aren't lazy scams. Fraud rings complete real coursework for a semester to stay eligible, steal the identities of real or former students, pose as homeless or underage applicants to dodge verification, and—increasingly—use deepfakes during Zoom identity checks. As LACCD Deputy Chancellor Nicole Albo-Lopez put it: "We learned we can't leave the back door open. Because they'll find a way in."

The Fix: N2N's LightLeap.AI

The central tool is LightLeap.AI, built by N2N Services. It screens applications at three points—when someone applies, when they register for courses, and when aid is disbursed—looking for the patterns a human reviewer can't see across millions of records: shared phones, emails, and addresses; device and IP fingerprints; mismatched time zones; suspicious age and course-taking patterns.

The system began at the Foothill-De Anza district, which signed a $56,250 contract with N2N in February 2024. Jory Hadsell—who built the model there and is now leading the systemwide rollout from the Chancellor's Office—explained why pattern-matching is the whole game: "When our colleges are looking record by record, they're not seeing patterns. But AI is really good at that."

Adoption has been rapid. By May 2026, roughly two-thirds of the 116 colleges were using or planning to use the tool, and the Chancellor's Office plans to embed LightLeap directly into CCCApply, the systemwide application portal, beginning summer 2026—so every applicant is screened at the point of application. (LightLeap is also deployed in Arizona, Michigan, and Minnesota.)

N2N reports strong results—though these are vendor figures, not independently audited: the company says LightLeap has prevented over $34 million in fraud since summer 2025 and identified 79,016 fraudulent applications out of 500,000+ screened systemwide. At Golden West College, which deployed the tool before spring 2025, President Meridith Randall estimates it eliminated roughly 96% of fraud—faculty went from flagging 10–15 fake students per class to two or three. N2N CEO Kiran Kodithala's framing: "The only answer for a bad guy with AI is a good guy with AI."

It's Not Just One Vendor

LightLeap is the biggest player, but the system is layering multiple tools:

ID.me is the identity-verification backstop, and the Board of Governors has made it mandatory for all students starting July 1, 2026. The rollout hasn't been frictionless: as of May 2026 only about half of students had completed verification, and ID.me's 18-and-over requirement complicates checks for the system's growing population of dual-enrolled high-schoolers. ID.me is careful to note its role is narrow—"only for identity verification… it does not approve admissions or financial aid."

The Catch: AI Catching AI Is an Arms Race

Two cautions belong in any honest version of this story.

First, the numbers are contested. In spring 2026, the U.S. Department of Education claimed its interim ID-verification guidance prevented $171 million in fraud in California. Both the Chancellor's Office and N2N pushed back hard—Hadsell said the federal policy changed little because Zoom verification was already happening, and Kodithala said flatly, "There's no basis for those numbers." Treat the $171M figure as a federal claim, not a settled fact.

Second, this is an arms race, not a victory. Every screening improvement trains the next generation of fraud. Scammers have already moved from fake addresses to deepfaked video interviews. Chris Ferguson, the system's executive vice chancellor of finance, is clear-eyed about the target: "The ultimate goal for our system is zero." The honest read is that AI has bent the curve—from ~$13M in a few months down to ~$500K a month—without ending the problem.

Why This Matters for Real Applicants

If you're a legitimate California community-college applicant, the practical upshot is simple: expect identity verification. Complete your ID.me check early (it's mandatory from July 1, 2026), apply with consistent personal information, and don't be alarmed if you're asked to confirm who you are—the friction exists because real aid was being stolen at scale.

It also reframes the broader debate. The headline fear is that AI is reading your essay. But the AI most likely to touch your community-college application first is the model deciding whether you're a real person at all. That's a different kind of AI in admissions than Georgia Tech's first-in-the-nation applicant AI policy or essay-scoring tools—and right now, in California, it's the one moving fastest.

---

Sources and Methodology

This piece draws on:

• California Community Colleges Chancellor's Office — Reimagine Apply (N2N as fraud-mitigation partner) and Digital Futures fraud-prevention pages (primary).
• ID.me — official California Community Colleges help-center documentation (primary).
• Reporting from EdSource (Sept 2025, May 2026), CalMatters (May 2026), Fortune (July 2025), Voice of San Diego (April 2025), and Local News Matters (May 2026), which attribute loss and fraud-rate figures to the Chancellor's Office.

Key limitation: the Chancellor's Office publishes fraud-prevention guidance but not a single consolidated, audited loss dataset; most quantified figures originate in its statements to journalists. Vendor-reported detection totals (N2N's $34M-prevented and 79,016-detected) are company claims.

At GradPilot, we help real applicants present their authentic selves—and we track how AI is reshaping admissions, for applicants and institutions alike.

Related Reading

• Which Colleges Use AI to Read Essays?
• Colleges Ban Student AI but Use AI to Read Your Essays
• What Universities Spend on AI Detection — $15M+ in Data
• Georgia Tech Wrote the First University AI Admissions Policy