Stanford's AI class: the fraud factory hiding in plain sight

by Luís Rijo 8 min read
Work Tech
Stanford's AI class: the fraud factory hiding in plain sight
Students write exams under a golden AI neural network looming above a vast exam hall.

One of Theo Baker's classmates summed up four years at Stanford in five words: "just a little bit of fraud." She was talking about unreturned sponsor hardware. Baker adopted the phrase as the thesis for his entire graduating class.

Baker is no ordinary aggrieved student. He arrived at Stanford in fall 2022, the same semester he broke the story that eventually forced the university's president to resign - work that earned him a George Polk Award, one of journalism's highest honors. Warner Brothers optioned the film rights. He is 21 years old. This week, he published How to Rule the World, a book about what Stanford actually teaches, and sat down with CBS News to explain what the class of 2026 - the first cohort to spend its entire college career alongside ChatGPT - discovered about AI, ambition, and institutional rot.

The answer is not reassuring.

The Background

To understand what happened at Stanford, you need to understand what Stanford was before ChatGPT arrived.

Stanford is not just a university. It is, by Baker's description in the CBS interview, "a place with a budget higher than 116 countries." Its endowment - the pot of invested money universities use to fund operations, similar to a sovereign wealth fund - sat at roughly $36 billion as of its last public filing. Its campus in Palo Alto sits at the geographic and cultural center of Silicon Valley, the cluster of technology companies that has produced more wealth per square mile than almost anywhere on Earth.

That proximity is not accidental. Stanford made a deliberate deal with the technology industry over several decades. It licensed its research cheaply, encouraged faculty to start companies, and allowed its brand to become effectively interchangeable with the idea of a technology career. In exchange, it received billions in donations, research contracts, and the kind of reputational capital that lets it charge students - and their families - accordingly.

The result, Baker argues, is that education became secondary. Stanford increasingly functions like a talent pipeline and deal network, not a place of learning. The cultural message absorbed by students was clear: the point is not to know things, it is to win. Getting rich, getting funded, getting hired - these are the metrics. The rest is decoration.

This culture pre-existed AI. The scandals did too. Baker reels them off: Elizabeth Holmes of Theranos, fraudster Do Kwon, the Juul founders. All Stanford alumni. All, in different ways, products of a culture that rewards moving fast and breaking things, a phrase that originated inside Silicon Valley and seeped back into the campus that helped build it.

Then, in November 2022, ChatGPT arrived.

What Is Actually Happening

ChatGPT is a large language model - software trained on vast amounts of text that can generate fluent, coherent writing on almost any topic, instantly, for free. It launched publicly two months after Baker's class began their freshman year. By the time they reached senior year, it had become, as Baker told CBS News, "this sort of massive accelerant" for every trend already present.

According to Baker's New York Times essay, AI has "permanently changed how we think and behave" for his entire graduating cohort. The cheating statistics Baker cites are remarkable. In his junior year, 49% of the 849 computer science majors who responded to an annual campus survey said they would rather cheat on an exam than fail. The surveys measuring admitted cheating, Baker argues, vastly undercount reality - nearly everyone, he says, is doing it some of the time.

Stanford's institutional response illustrates exactly how serious the problem became. In April 2026, Stanford's Faculty Senate unanimously voted to authorize proctored in-person examinations - ending a policy of trusting students that had been in place since 1921. More than a century of institutional faith, dissolved in a single vote. Most exams are now handwritten in Blue Books, the paper exam booklets that students used before personal computers existed. The practice had experienced a significant resurgence across American universities since 2024 as AI cheating spread.

The plagiarism went beyond homework. According to The Decoder, Baker details a case where Stanford students published a paper claiming an AI breakthrough with a model called Llama3-V - which turned out to be a stolen Chinese model. They had taken someone else's work, rebranded it, and submitted it for academic credit.

The stratification is equally stark. Baker notes that CS enrollment at Stanford declined last year for the first time in over two decades - a signal that students themselves are recalibrating the value of a traditional computer science degree. Meanwhile, a classmate who started an AI company six months into freshman year saw it valued at over a billion dollars. The distance between those two outcomes - dropout billionaire and unemployed graduate - is the defining economic story of Baker's cohort.

The Money Trail

Follow the money and the picture sharpens considerably.

The incentive structure at Stanford has been distorted by the same forces that distorted the rest of Silicon Valley. Venture capital - the money professional investors put into early-stage companies in exchange for a share of future profits - has flooded into anything with an AI connection. According to The Decoder, so-called wrapper startups - companies that do little more than repackage existing AI models with a slightly different interface - were attracting enormous valuations. Perplexity, a search tool built largely on other companies' models, hit a $1 billion valuation in April 2024 and reached $20 billion by September 2025.

The message this sends to students is not subtle. The path to wealth does not require original research, deep expertise, or even a completed degree. It requires a company name, an AI suffix, and the right network. Baker's dropout classmate understood this before most professors did.

At the same time, the traditional payoff for a Stanford CS degree has collapsed. A Stanford professor told the Los Angeles Times that graduates are "struggling to find entry-level jobs" - a dramatic reversal from just three years ago. According to Federal Reserve Bank of New York data, CS graduates now carry a 6.1% unemployment rate in 2026, nearly double the rate of philosophy majors, while entry-level software engineering roles dropped roughly 30% year-over-year. Junior developers now compete directly with language models that can write functional code faster, cheaper, and without benefits.

This creates a rational, if disturbing, economic calculation. If the expected value of doing the work honestly has dropped - because the job it was supposed to unlock may not exist - and the expected value of cheating has risen - because AI makes detection harder and the penalties remain low - then cheating becomes the economically optimal choice. Baker is not excusing it. He is describing it.

Stanford profits from both ends of this arrangement. It charges full tuition to students whose degrees have declining market value. It licenses its research and brand to the tech industry that is simultaneously destroying its graduates' job prospects. The endowment keeps growing. The students bear the risk.

The Faustian bargain Baker describes - the deal Stanford made with Silicon Valley that brought billions in exchange for a "move fast and break things" culture on campus - has now produced a generation that learned to move fast and break the honor code.

What People Are Doing About It

The institutional responses are arriving, though slowly and mostly in the form of reversals.

Stanford's proctoring policy represents the most visible change. Beginning fall 2026, instructors are permitted to proctor any in-person assessment. The Faculty Senate, Undergraduate Senate, and Graduate Student Council all voted in favor - a rare display of institutional consensus. The Academic Integrity Working Group that drove the change was itself formed in 2024, after updates to the Honor Code following a documented spike in violations.

Beyond Stanford, other elite institutions are following the same path. Princeton, which similarly relied on an honor system, is undergoing comparable rule changes. The Wall Street Journal reported that blue book exams have experienced a significant resurgence across American higher education since 2024, as professors scramble to find assessment methods AI cannot easily infiltrate.

Students themselves are shifting. According to the San Francisco Chronicle, CS enrollment dropped 6% system-wide across University of California campuses in the 2025 academic year, after a 3% decline in 2024. National CS enrollment fell 8.1% in the 2025-2026 school year, according to National Student Clearinghouse data, with students shifting toward AI-specific programs, data science, and cybersecurity. The one UC campus that bucked the trend - UC San Diego - was the only one to add a dedicated AI major.

Some commencement speakers are getting booed. At the University of Arizona this month, former Google CEO Eric Schmidt was met with audible disapproval when he celebrated AI's potential during the graduation ceremony. According to The Verge, students who spent four years being told AI was the future are not especially enthusiastic about being told, on graduation day, that it will make their lives better.

Baker himself, asked whether his reporting tarnishes his own diploma, gave the careful answer of someone who has spent four years watching how institutions protect themselves. Stanford, he said, is "a complicated thing." There are great professors, great students, great innovations. And there is also the other thing.

The Bottom Line

AI did not corrupt Stanford. It found a culture already primed for corruption and gave it better tools. The same technology that minted a dropout billionaire from Baker's freshman dorm is the one that eliminated the entry-level jobs his classmates spent four years and six figures preparing for. The people who got rich were mostly already rich, already connected, already inside the network. The people who cheated on exams got a credential of declining value. Stanford collected tuition either way. That is the economic logic of the class of 2026, written in Blue Books by hand, under supervision, for the first time since 1921.

Timeline

Summary

Who: Theo Baker, a graduating Stanford senior, investigative journalist, and author, representing the broader experience of Stanford's class of 2026.

What: AI arrived at Stanford two months after the class of 2026 began, accelerating an existing culture of academic fraud, widening wealth inequality between students, and contributing to the collapse of entry-level CS job prospects - prompting Stanford to end over a century of unproctored exams.

When: The period from fall 2022, when ChatGPT launched, through May 2026, as the class approaches graduation.

Where: Stanford University in Palo Alto, California, at the center of Silicon Valley.

Why: AI functioned as an accelerant on incentives already warped by Stanford's financial entanglement with the technology industry - making cheating easier and cheaper while simultaneously destroying the job market the degree was meant to unlock.

Luís Rijo

Luís Rijo

Luis Rijo has spent over a decade watching money move through digital markets - where it goes, who follows it, and who gets left behind when the flow changes direction. [email protected]