Advertisement

The University of Cincinnati and the tech firm Nexigen are working to improve the safety and security of AI systems.

Digital generated image of artificial intelligence yellow brain on cell patterned blue background with geometrical shapes on it.

The University of Cincinnati and the tech firm Nexigen last week announced a new collaboration that aims to make artificial intelligence more secure.

The partnership builds on Nexigen’s work to create secure repositories outside of large language models that can generate content by processing large amounts of data. Experience from the university’s faculty will help to create more filters and increase the accuracy and safety of the secure repositories. 

“I feel there’s a lot of value to our students and new opportunities for research,” Kelly Cohen, chair of the aerospace engineering department at the university, said in a press release. “We’re trying to establish ourselves at UC as a powerhouse when it comes to AI and trustworthy AI. You build this not by doing something alone in an ivory tower by bringing value to your partners.” 

The new partnership also seeks to improve the accuracy of AI output and eliminate ways that people can circumvent guardrails to get instructions for illegal activities, such as building bombs. 

Advertisement

Jon Salisbury, Nexigen’s chief executive, said that as AI software becomes more advanced, it could present threats of its own and that the partnership also intends to tackle those challenges. 

“We take a cybersecurity approach, and it’s zero trust,” Salisbury said in the release. “So we are looking for potential intent problems with the output on LLMs so that if they were to develop a sense of self intelligence, we can actually detect that and kick them off the network.”

Skylar Rispens

Written by Skylar Rispens

Skylar Rispens is a reporter for StateScoop and EdScoop. She previously worked as a reporter specializing in education coverage for daily and weekly newspapers across Montana, where she currently resides.

Latest Podcasts

Higher Education

Cybersecurity

Hybrid Learning