Julia Angwin

Julia Angwin is an American investigative journalist who has spent two decades holding technology to account. A Wall Street Journal reporter from 2000 to 2013, she shared the 2003 Pulitzer Prize for Explanatory Reporting for coverage of corporate scandals, then moved to ProPublica, where her 2016 “Machine Bias” investigation showed that the COMPAS criminal risk-assessment algorithm was biased against Black defendants — the story that effectively founded the algorithmic-accountability beat. In 2018 she co-founded The Markup, a newsroom dedicated to investigating big tech, and later launched the nonprofit studio Proof News.

Auditing the AI boom

In July 2024, Proof News published an investigation showing that subtitles from 173,536 YouTube videos across more than 48,000 channels — swept up in the “YouTube Subtitles” portion of the open Pile dataset — had been used to train AI models at Apple, Nvidia, Anthropic, and Salesforce, without the knowledge of creators ranging from Khan Academy and MIT to Marques Brownlee and MrBeast. The story became a touchstone in the fight over consent and training data.

That May, in a New York Times op-ed titled “Press Pause on the Silicon Valley Hype Machine,” Angwin argued the debate had inverted: the pressing question was no longer whether AI would grow too powerful, but whether it was too unreliable to be useful — a “decent first draft” machine whose output takes nearly as long to fix as the work it replaces, built by companies running short of both training data and energy.

Angwin’s stance is that of an empiricist, not an activist: she tests claims, measures systems, and publishes what she finds. But the through-line of her findings — biased algorithms, unconsented data extraction, overstated capabilities — amounts to one of journalism’s most sustained public warnings about the AI industry.