A year ago, USC announced the release of ChatGPT Edu for students, faculty and staff, and has since introduced various classes, majors and the Stevens School of Computing and Artificial Intelligence. In that same time, USC has been working toward a sustainability deadline of its own: under its Assignment: Earth framework, the University set a goal of reaching climate neutrality for annual Scope 3 emissions — air traveling, commuting, waste, power transmission and distribution losses — for fiscal year 2035 and beyond.

Mick Dalrymple, USC’s former chief sustainability officer, said that as of right now it seems like AI is not a large portion of the University’s footprint. But he added, it’s “very hard” to measure the impacts of University-AI usage.

“Right now, there are no industry standards on how data centers and AI companies report stuff,” Dalrymple said. “What we’re trying to do now is we’re working very closely with ITS, which is the … Information Technology Services, in trying to get our vendors to disclose their emissions, their water use, things like that.”

Amit Jain, adjunct assistant professor of sustainability management, said one of the hurdles of calculating the environmental impacts of AI is figuring out how to categorize the emissions as well as how to calculate the environmental impact of a tool whose operations are so far removed from the institution utilizing them.

The Office of Sustainability is working closely with the University’s Information Technology Services to get vendors to disclose their emissions data, but it remains difficult without any industry standards for gathering emissions data, Dalrymple said. 

“We’re trying to figure out is there another way that we could calculate it, but we’re somewhat dependent upon them to to give us something and then tell us how they calculated it, so that we could say, ‘Oh, we agree to that,’ or ‘We don’t agree with that,’” Dalrymple said.

OpenAI, who is currently partnering with USC on the educational front, is one vendor in particular that doesn’t disclose such customer-specific information.

In 2024, The Washington Post estimated the GPT-3 version of OpenAI’s ChatGPT uses around 500 milliliters of water for each short conversation. Google has also reported the median single user text query uses around 0.24 watt-hours of energy, 0.03 grams of CO2 equivalent and 0.26 milliliters of water for cooling.

Dalrymple added that right now there’s a “gold rush mentality” around AI in general.

“With the pace at which this is happening, the environmental aspects and the social impacts are not given as much concern as they should,” Dalrymple said.

Since the University began implementing AI in the classroom, faculty have voiced concerns over how such programs may affect students’ critical thinking skills. Essentially, they felt the new AI tools run the risk of “deskilling” students. A 2026 University survey found that the about 1,000 respondents averaged between somewhat concerned and very concerned regarding chatbots impacting academic integrity. 

With a lack of transparency into the environmental impact of AI, some students and faculty said they are still undecided about the trade-off between benefits and harms.

As an environmental, social and governmental executive, Jain remains torn. He recognized there are useful tools AI can provide, such as aggregating environmental data and gathering its analytics that would otherwise be costly, while also acknowledging the need for more discussion about its environmental costs.

“USC has financial control over people using [AI], but they don’t have operational control over how people use it [and] when people use it … but they should definitely try to seek that out,” Jain said.

Shannon Gibson, professor of environmental studies, political science and international relations, said that as director of the Gibson Climate Justice Lab, she observed partnering organizations’ apprehension toward working with a University that is “very much supporting AI use in higher education” amid the pushback against data centers and AI expansion.

“It’s important to avoid flattening the debate into [whether] AI is good or AI is bad, and to acknowledge the tension that exists because there are many drawbacks and concerns, but there are also benefits that should be considered,” Gibson said.

While the initial launch of AI tools at the University was met with pushback from some faculty, some students are still grappling with what the future of AI entails.

“From an engineering standpoint … no matter what you think about AI, it’s a part of the world right now … you’re going to use AI in some fashion,” said Isabel Ketterson, a junior majoring in environmental engineering. “We have to focus more on educating people about when and how to use it correctly, to use it sparingly, as well as to put effort into researching ways to make it more sustainable.”

In the meantime, Dalrymple said USC is working with other universities to consolidate their knowledge and better educate themselves.

“From a sustainability standpoint, we’re definitely working together to try to figure this out very quickly, and we also want to be able to track it in a uniform way ourselves,” Dalrymple said.