Smart, Confident, and Wrong: Designing Responsible A.I. Tools in the Newsroom

Designing AI tools in a newsroom means working with systems that sound confident even when they are wrong, inside a profession built on accuracy and trust. This session shows how careful product and workflow design helps newsrooms capture the benefits of AI without sacrificing editorial standards. Drawing on tools built at The New York Times, the speakers share practical design principles, demos, and industry examples that focus on constraining hallucinations, tracing sources, and building AI products journalists and audiences can trust.

// speakers

Dylan Freedman

A.I. Projects Editor

The New York Times

LinkedIn →

Dylan Freedman is the A.I. Projects Editor at The New York Times. He reports and builds investigative tools to help journalists make sense of a wide range of topics. Before joining The Times, Dylan worked on a machine-learning team at Google, served as the lead developer for the journalism nonprofit DocumentCloud and architected elections platforms as a principal software engineer for The Washington Post. He has taught data journalism courses at Temple University and University of Maryland, College Park. Dylan graduated from Harvard with a bachelor’s in computer science and music and from Stanford with a master’s in journalism. He lives in Washington, D.C.

James O'Toole

Senior Machine Learning Engineer, A.I. Initiatives

The New York Times

LinkedIn →

I'm an engineer and journalist on the A.I. Initiatives team at The New York Times, where I use A.I. for newsroom tooling, reporting projects, and reader-facing prototypes. I previously worked on the newsroom engineering team at The Washington Post, and began my career as a reporter for The Phnom Penh Post in Cambodia before covering finance and technology for CNN Business. I've also worked in startups, local government, and academia.