Which AI risks end up in legislation, and where does that ranking come from?

Nick Diakopoulos has spent a decade on that question. He is a professor of communication studies and, by courtesy, computer science at Northwestern University, where he directs the Computational Journalism Lab.

He put the phrase "algorithmic accountability reporting" into circulation in a 2013 essay for The Atlantic and a 2014 Tow Center report, and his book Automating the News: How Algorithms Are Rewriting the Media (Harvard University Press, 2019) won the Tankard Book Award.

In this conversation he walks through two studies. The first analyzed 42,853 news articles across 27 countries to measure which AI harms actually get covered: societal risk dominates at roughly half of all coverage, legal and rights harms take about a third, and existential risk finishes fifth at 7%.

The second, presented at AAAI/ACM's AI, Ethics, and Society conference in 2025, narrows to US outlets and finds that risk coverage sorts by political lean — right-leaning outlets emphasizing malicious use, left-leaning outlets emphasizing socioeconomic and environmental harm, and both using the word "discrimination" to describe different phenomena.

The conclusion is uncomfortable and hard to dismiss: AI risk assessment inherits the politics of the information reaching the people who write the rules.