USC launched their ChatGPT Edu workspace at the start of the spring semester, previously announced at the inaugural artificial intelligence summit in November, which can now be accessed by active students, staff and faculty. The launch also coincided with new generative AI policies.

However, some faculty call the partnership “untransparent,” citing no communication of the cost of the institutional subscription after a semester filled with faculty layoffs. Some faculty have also raised concerns about the potential costs, considering that in the Fall 2025 semester USC laid off more than 1,000 faculty and staff.

When asked about the cost of the USC’s partnership with OpenAI, interim Chief Information Officer Midhat Asghar wrote to the Daily Trojan that the agreement is a one-year subscription that was negotiated with “a strong focus on financial stewardship and value for the university.”

“When compared to other enterprise-grade AI offerings in this space, the cost is the best value for the university and meets USC’s requirements for security, privacy, and scalability for a campus-wide deployment,” Asghar wrote. 

User input will not train OpenAI models

On a FAQ page about USC’s ChatGPT Edu, the University wrote that it adopted the chatbot because many community members were already incorporating it. It also stated that the Edu workspace will allow many to use the chatbot “without training OpenAI’s models.” 

In the University statement to the Daily Trojan, Asghar also wrote that the University received multiple requests from faculty, staff and students for an enterprise-grade AI tool. Asghar also stated that ChatGPT had about 40,000 unique monthly users using the @usc.edu email before the enterprise launch. 

The FAQ page also stated that USC-supported AI tools are optional and instructors may prohibit chatbot usage. If usage is permitted however, students are told to cite its use in their coursework. 

The website also recommends ChatGPT users to never use real names, IDs or confidential info in their messages with ChatGPT, and instead use generic or fictional data for data privacy. USC’s ChatGPT Edu workplace gives a personalization setting to enter the user’s preferred name, occupation and additional information for the chatbot to incorporate into responses. 

The University issued a new generative AI policy on Jan. 9, stating  that it is the user’s responsibility to check if the chatbot’s responses are accurate. Generative AI chatbots, like ChatGPT, are known to hallucinate, meaning it confidently responds with inaccurate or nonsensical information. 

The new generative AI policies also prohibits students, faculty and staff from entering confidential data such as medical data or sponsor-restricted data into both individual chatbots and ChatGPT Edu. However, the new policy allows for internal data, like student IDs, to be typed into ChatGPT Edu but not into other chatbots outside of USC’s workspace. 

A violation of the new policies may be classified as serious misconduct and can lead to disciplinary action or termination, according to the site. 

When USC’s ChatGPT Edu workplace is opened, a window appears, reiterating that typing into the workplace’s chatbot will not be used to train OpenAI’s models nor will any messages be monitored by the University. The workplace also disables third-party GPTs, except for approved models.  

Users can change between different models, which specialize in different functions, including the latest model, GPT-5.2, as well as legacy models GPT-5.1, GPT-5 and GPT-4o.

Some faculty push back

A letter signed by 12 USC professors was sent to the Daily Trojan in November, critiquing the University’s institutional subscription to ChatGPT and the preceding announcement at the AI Summit. 

“USC has told students it can’t afford to pay the real people they trusted. Instead, it’s buying them a pretty toy,” the letter read.

Three professors who signed the letter talked to the Daily Trojan, sharing criticism against the subscription and the University’s lack of transparency about the deal. 

Aro Velmet, an associate professor of history, called USC and OpenAI’s partnership as an “unthought-through collaboration” that is untransparent and possibly expensive. Velmet also said that this subscription “generates a series of labor issues.” 

“Faculty now have to spend time figuring out what this means for their classes, figuring out whether they want to use this or don’t want to use this,” Velmet said. “This is all time that faculty could be spending actually teaching students and meeting with them and giving them individualized guidance.”

Kate Levin, an associate teaching professor of writing who signed the letter, said that the majority of faculty were not consulted, polled or informed about how the institutional subscription to ChatGPT would affect their classes prior to the November announcement. 

“If we’re not being consulted about a decision like this that not only impacts, but potentially jeopardizes, our ability to do our job, it raises really serious concerns about how the University is being run, especially since they really do love to tout this idea of shared governance,” Levin said.

Patti Taylor, an associate teaching professor of writing who co-authored the letter, said she used to be very interested in what generative AI could do and joined the Center for AI in Society because of it. But overtime, after learning of its capabilities, became less excited about these chatbots and now believes it could be damaging to learning. 

Specifically, Taylor cited research into generative AI that showed a strong connection between confidence and lower generative AI usage. She also cited emerging research on “deskilling,” where a doctor’s skill in detecting cancer cells dropped after heavy AI usage in about three months. 

“That’s the kind of thing that really worries me, in terms of handing ChatGPT to students and just saying, ‘Use it whenever,’ without having a lot of guardrails saying, ‘Well, are we bothering to train our students on when they should and shouldn’t use it?’” Taylor said. 

Levin said the classes centered on writing and critical thinking are still contending with ChatGPT usage, which does the thinking, brainstorming and writing — all practices used to develop student’s critical thinking. 

“Now, if the institution itself is coming along and saying, ‘You have not just our blessing but our encouragement to use this tool,’ without consulting any of us who are doing the on-the-ground work, that’s concerning,” Levin said.

Last semester, Velmet, who specializes in science and technology history, said his history class on science, technology and medicine had a large chunk dedicated to the history of AI. In his class, he makes a point to say that generative AI models — which he prefers to call “chatbots” — are neither artificial nor intelligent. 

“It’s not artificial, in the sense that, it relies on a tremendous amount of human labor,” Velmet said. “It’s not intelligent. It produces grammatically correct text based on a statistical model. You talk to anybody in cognitive neuroscience and nobody thinks that that’s intelligence.”