SALT LAKE CITY (ABC4) — In recent weeks, international students, families and naturalized citizens across the U.S. have been informed that their immigration status had changed or that they have mere days to leave the country.

One attorney has recently told affiliate KTVX that new technology like AI may be behind mistaken visa revocations or emails demanding self-deportation. KTVX spoke with a local expert about what role tech might be playing in the Trump administration’s immigration crackdown.

Sharad Jones is a professional practice assistant professor at Utah State University. He said government employees who are dealing with immigration have large caseloads and face “constant changing political pressures.”

“We need to identify places where AI — kind of automation technologies — allow us to speed up the necessary pieces of our workflow, but don’t replace the decision-maker… the person that is held accountable for that final piece of it. That’s not an easy thing,” Jones said.

Jones said that there are several data points and patterns that an algorithm would need to look at to determine a person’s citizenship status. However, he said one limitation is that “the only way to know for sure” whether someone is a resident or not is to speak to a person directly — but that is difficult to do on such a large scale.

“There’s not been a single machine learning model that has been developed that is perfect and has never made an error,” Jones said. “To me, this sounds like one of those cases… imperfect data resulted in an imperfect model, which therefore can result into in-error cases, and this happens every day.”

He said that it is important for technology to improve efficiency, but it is also important not to reinforce issues in the data: “These aren’t unknown problems. This is us rehashing the same things we keep doing.”

Attorney Adam Crayk has previously speculated that new technology is being used by the Trump administration to enforce new immigration policies — and mistaken visa revocations or self-deport emails may be a result of faulty technology.

“Technology has its place. It does,” Crayk told KTVX. “But we can’t let it run rampant. We can’t let it make every decision for us. We actually have to intervene at some point and look and see if it’s making correct decisions.”

Jones encouraged people to learn more about technology as it develops, and said a lot of concerns with AI are reminiscent of concerns with the internet when it was first becoming mainstream.

“The best bit of advice I can give to anybody is to try and learn just a little bit about this technology. Really try to understand at its core what it’s trying to answer,” Jones said.

Matthew Drachman contributed to this report.