The CSU paid millions for every student to have ChatGPT. What happened after?

Josue Cruz, an incoming junior at San Francisco State University, knew something had changed when his linear algebra professor urged the class to use artificial intelligence — not to cheat by using a chatbot to do their homework but to ask it questions about their assignments.

The professor encouraged students to use ChatGPT as a study partner, Cruz told SFGATE, that could explain a difficult equation and walk them through a problem. When the time came for exams, though, students were asked to put their laptops away and pick up their pencils. For Cruz, who’s studying computer science and fears the technology could potentially take a future job, it was startling to hear a professor endorse AI as part of the learning process.

“She encouraged AI but as a tutor buddy, and that took me by surprise,” Cruz said. “… We didn’t have the chance to rely on AI during the test or during homework. It was just all paper.”

Article continues below this ad

If you went to college a decade ago, the scene might sound dystopian. In just a couple of years, AI has radically changed learning to the point where it can now draft an essay, solve a math problem, write computer code or build a study guide before students have even begun typing the question. Some students have even used it to cheat, letting the chatbot do the work for them while they take the credit.

But as campuses across the California State University cut courses, lay off faculty and struggle with plunging enrollment, the system hasn’t just approved the use of OpenAI’s ChatGPT as an equitable and advanced learning tool; it’s also investing in it. After spending $17 million on a one-year OpenAI contract, the CSU renewed the partnership last spring for three more years at $13 million per year. Though the deal makes AI tools and training available to more than 460,000 students and 63,000 faculty and staff across all 23 campuses, it’s also drawing fierce opposition over the technology’s hefty cost, its effectiveness and its impact on learning. Even at San Francisco State in the “AI capital of the world,” faculty members and students told SFGATE that the rollout has produced a more complicated reality: inconsistent usage rules, limited training, questions about academic integrity and disagreement over whether giving every student access to a chatbot actually helps them learn.

“The initiative raises serious questions about classroom practice,” Martha Lincoln, an associate professor of anthropology, told SFGATE. “Because it’s essentially impossible to craft policy around AI use or prevention, San Francisco State cannot mandate the use of AI. But they also cannot ban it.”

The first year of the rollout: Was it worth it?

When the partnership was first announced in February 2025, OpenAI also described the deal as the “largest implementation of ChatGPT by any single organization or company anywhere in the world.” Meanwhile university officials said the partnership, which gives students, faculty and staff access to ChatGPT Edu across its campuses, would help close the “digital divide” and prepare students for an AI-shaped workforce. Though ChatGPT has a free version of its software, the CSU’s investment, ChatGPT Edu, does not use data to train its large language models, and according to CSU’s FAQ page, it has “advanced capabilities” such as data analytics, web browsing, higher message limits and document summarization.

“The CSU is recognizing that AI is a part of our world and that the students will have learned when they come to our institutions to understand what it is and then also because we’ve got all these trainings, learn responsible use cases of it,” Leslie Kennedy, the CSU’s assistant vice chancellor of academic technology services, told SFGATE.

ChatGPT website displayed on a laptop screen is seen in this illustration photo taken in Krakow, Poland, on April 16, 2026.

In the first year of the university system’s $17 million partnership with OpenAI, Kennedy said people were signing up for ChatGPT Edu slower than CSU officials expected. By the end of the spring 2026 semester, at least 250,000 accounts were activated throughout the system, Stephanie Metzinger, a spokesperson for the CSU, confirmed to SFGATE. With the new agreement, students are also able to use the technology up to one year after graduation, Metzinger said.

“The pro was that now everybody had a tool, so we did not have to worry about the digital divide,” Anoshua Chaudhuri, an economics professor at San Francisco State who has helped lead the campus’s work on artificial intelligence, told SFGATE. “The con was that now everybody had to sit up and take notice.”

But getting AI into the CSU was only the first step; faculty members said the bigger challenge is making sure people know that it’s there and what to do with it. To ease the introduction, the CSU system and the individual campuses have launched training programs, including self-paced courses and seminars that are available to students, staff, faculty and alumni. For example, Academic Technology at San Francisco State even developed an “AI literacy” course for the entire university system, which is an optional Canvas course that takes three to four hours to complete.

“We work closely with our sister campuses across the CSU and the office,” Andrew Roderick, the assistant vice president of academic technology, told SFGATE. “I do think there’s a lot of disparities with how campuses are able to build up programming and support for their faculty and students around AI.”

But even with those programs, students aren’t flocking to technology. Similar to the systemwide trend, Roderick said the rollout at SF State has also shown low usage, with only about 22% of students, faculty or staff activating the software. Some students are also “distrusting” of the AI software provided by the university because they feel like university officials will monitor what they’re doing.

“Many of them use their own tools and in some cases their own tools out of their own preferences,” he said. “We know many of our students have concerns about AI technology in general and the negative impacts attributed to it.”

Travel | French tourist dies in Death Valley after car breaks downFood | This is the most polarizing restaurant in the Bay AreaNews | Jan. 6 attendee sells Bay Area cafe, says money goes to MamdaniCrime | Arrest made 18 years after SF man killed on eve of his wedding

Read our stories ad-free on the new SFGATE app.

So far, Cruz said, he’s heard the same sentiment from his peers. He said some students who don’t use the school’s provided system are worried about privacy concerns such as SF State or CSU officials being able to read up on their chats.

“Even though they say they’re not training any of the data, there’s always concerns about like some of the chat logs leaking,” he said.

Chaudhuri said “anecdotal evidence” from faculty members suggests that most students who do use AI are mainly using their personal accounts for classwork. Although faculty groups and university officials have offered extensive training and resources, including an “AI readiness” toolkit at San Francisco State, which Chaudhuri described as a starting point for those unsure where to begin, some students and professors still feel in the dark.

Some students who’ve pushed back on the rollout have cited environmental concerns, including the significant amount of water and electricity that centers need to power the software.

Student feelings and mixed messages

One major point of tension is that the CSU system hasn’t set one clear policy for all classrooms. Instead, its FAQ says faculty “retain full authority over their classrooms,” meaning they are able to continue pursuing their own teaching methods and syllabi and that using AI is voluntary. But that ambiguity has produced an uneven landscape for students.

Martha Kenney, a San Francisco State professor who helped lead a petition against the contract’s renewal, called the first rollout “minimalist,” because the CSU system wasn’t really pushing the technology in a particular way. She said she took the school’s AI literacy course, which she described as “very basic” because it didn’t focus on specific applications for teaching or learning.

“Students are getting mixed messages about what this technology is, what it’s for, what you can use it for and what you can’t,” Kenney said.

She told SFGATE that students are receiving different responses from different instructors about what ChatGPT is and whether it is considered legitimate in college work. Some students also don’t know that the partnership exists or that they have access to the Edu chatbot.

“I don’t think there’s much advertisement about it. At least from what I’ve seen,” Cruz said. “… I’d like to see more transparency like, ‘OK, we bought the deal,’ but what is offered for us as the students?”

Kenney said the technology became “really hard to ignore” starting in 2024, even before the contract existed. That’s when she began seeing “work that shows hallmarks of being written by chatbots,” prompting her to talk with students about it. Since then, it’s been challenging to set standards for using ChatGPT and other chatbots for students.

Exterior of San Francisco State University.

“It would already be a challenge, but when the university actually provides the technology, it adds kind of an additional layer of legitimacy,” she said. “… I’d say 2024 was kind of flexible and where it seems like OK, this is actually going to change how we’re able to do our job as educators.”

According to the CSU’s FAQ page, officials see AI as becoming essential to the workforce. The CSU’s first systemwide survey on AI, published in April, just a month before the OpenAI deal was renewed, revealed that AI use is widespread and that students and faculty recognize that reality, even as they remain concerned about accuracy, ethics and job security.

According to the survey, which was spearheaded by San Diego State University with responses from more than 94,000 students, faculty and staff across the system from fall 2023 to fall 2025, 69% of students believe AI will be essential in their jobs, but 82% are worried about job security. Only 38% of student respondents believe that the AI information is accurate, and 76% are concerned about ethical use.

‘Not an educational technology’

Lincoln said the CSU partnership reveals a “totally explicit message that AI has a place in higher education.” She cautions, though, that “we are not a vocational training institution.”

Faculty critics of the deal still say the CSU has emphasized access and responsible use without answering a more basic question: Does ChatGPT actually improve students’ education?

If students are using the technology for tutoring or quizzing, for example, Kenney said she’s worried that it’s unclear if the information the chatbot is giving students is accurate. One of her main concerns is not that students have the technology but that the chatbots aren’t trained specifically as an educational tool.

“Equity doesn’t matter if it’s not an educational technology,” Kenney said. “If I gave every CSU student a can of Coke, that’s equitable. But is that going to help their education?”

In a 2025 study at the University of Minnesota and Dublin City University, researchers analyzed some college classrooms where AI was openly used. In one case, a professor asked his students to use ChatGPT to assist them with their essays; however, some students accepted the output without verifying the information. Many of the essays had “made-up sources” and wrong information, the study states. And in another professor’s physics class, some students who used ChatGPT also blindly followed the chatbots.

“… Students believed that ChatGPT had superhuman intelligence and was practically infallible, even when it answered their physics questions incorrectly,” the study says.

If students are misled by ChatGPT’s information, Kenney and other professors believe it would do more harm than good.

“We’re not really concerned about the cheating,” she said. “We’re concerned that the CSU is cheating our students out of a good education, and that’s like one thing that the CSU has historically been able to provide.”

In Lincoln’s classroom, she said the technology has changed the “traditional process of completing work.” Now many teachers are forced to create assignments that are supposedly AI proof or AI resistant, but Lincoln said she questions whether those assignments truly benefit students.

For example, Lincoln said she doesn’t assign research papers anymore, despite it being a skill that most of her students want to learn.

“I can’t assign that anymore because CSU administration essentially greenlit the use of AI by students with very few parameters around that use and no coherent policy,” she said. “… It’s like everybody’s hands are tied, and I can’t trust my students in the normal way that’s healthy for them and healthy for me.”

A student looks toward a plywood box covering a statue of Cesar Chavez at Cal State Fresno in Fresno, Calif., Wednesday, March 18, 2026.

Concerns about how chatbots could potentially misguide students also come at a time where students across the country are already struggling with their academics because of the pandemic. After several years of isolation and spending formative high school years in remote learning, the current generation of college students is largely coming into universities underprepared and with lower test scores, SFGATE previously reported. The CSU has also acknowledged the “unfinished” learning from the pandemic that is negatively affecting its current students.

Lincoln said she was “disturbed” by the administration’s making the decision and believes officials should’ve asked for more input from faculty and students.

“It’s really undemocratic,” she said. “It’s a bad look.”

In the petition launched earlier this year by Kenney, Lincoln and other CSU faculty members, opponents of the partnership urged CSU Chancellor Mildred Garcia not to renew the contract and called for the money to be redirected toward faculty, staff and student resources. Titled “Cancel ChatGPT Edu. Invest in Humans,” the petition, which had 4,072 signatures as of Friday, argues that it’s not worth the money to invest in a general-purpose chatbot that “does not use reliable peer-reviewed sources” when there is already a free version of ChatGPT.

The university system also is facing a $2.3 billion budget deficit, resulting in layoffs and extensive cuts. Some campuses, including San Francisco State University, Cal State East Bay and Sonoma State University, have also been forced to merge some operational services in recent years to save money amid falling enrollment numbers.

Opponents of the partnership said the budget crisis and crippling enrollment is one of the main reasons they oppose it.

“For instance, we don’t have janitors who take our trash out of our office anymore. We have to bring our own trash to a central location,” Kenney said. “We’re really bare bones at this point. The whole department budget for women and gender studies for a year was like $1,200.”

Faculty members also questioned the lack of publicly stated accountability measures over whether AI use has improved student performance or harmed their critical-thinking skills.

“Here they are giving Silicon Valley like $17 million for a technology that you can get for free online that has not at all been designed or trained or customized for education at all,” Kenney said. “… There are no accountability mechanisms on this. They’re not rating whether this contract is successful at all.”

According to Kenney, members sent the signatures to CSU administrators twice, but they did not receive a response.

The CSU now has three years to show whether its partnership with OpenAI can deliver more than broad access to a popular technology.

Kennedy said the system got the “best price in the country” for the partnership. And going forward, there will be more opportunities for training, and the university system will start to figure out more complex ways people can begin using the software, rather than just asking it “what to have for dinner tonight,” she added.

“There’s always opportunities with a half a million students and 28,000 faculty and about the same number of staff to continue the discussion in as many venues as possible to share what’s available,” she said. “… I think more and more we’d like to see a deeper institutional integration.”

Roderick said he’s beginning to see more faculty at SF State that are being open to talking about AI or using it in the classroom as some, along with students, are recognizing it as “very important for their career.”

Students walking to class on the CSU Northridge campus.

“It’s going to be up to them to make really great decisions around how it is used in our economy, in our society, and in their day-to-day lives,” he said. “… It feels like what we need to be doing is educating and training our students on how to think about the technology.”

But for others, the renewal is not simply a technology decision. It is a decision about what the CSU believes a college education should provide at a time when students are already navigating financial pressure, pandemic-related learning gaps and a changing job market.

“I’m kind of scared for the future,” Cruz said. “… I think especially within [the computer science] degree, AI is definitely taking over jobs. That’s a major concern. The school needs to focus on what AI cannot target.”

— Brewery to lay off over 200 as it shuts down Calif. headquarters— French tourist dies in Death Valley after car breaks down— Bay Area man arrested on suspicion of killing 4 family members— Jan. 6 attendee sells Bay Area cafe, says the money goes to Mamdani

Sign up for SFGATE's breaking news email