Earlier this week, the Washington Post announced that it would be launching “personalized” AI powered podcasts that would let users choose their own AI host to regale them on their choice of topics.
And now for an entirely unsurprising update: the AI podcasts have turned out to be complete, error ridden disasters.
Semafor reports that less than 48 hours after launching, the AI podcasts have sparked outrage among the WaPo’s rank and file and editors alike, after they caught the AI-generated podcasts committing ghastly journalistic sins, like inventing quotes and misattributing information.
“It is truly astonishing that this was allowed to go forward at all,” one WaPo editor fumed on Slack. “Never would I have imagined that the Washington Post would deliberately warp its own journalism and then push these errors out to our audience at scale.”
“If we were serious we would pull this tool immediately,” the editor added.
The podcast’s errors are exactly the kind you’d expect an AI model to make. Some are simple but noticeable cases of mispronunciation. But at times, according to Semafor, the AI podcast hosts would insert commentary, essentially editorializing by misconstruing a source’s quote as the paper’s position on the issue.
In Slack messages obtained by Status, other staffers railed blasted the AI feature.
“What are the guardrails to ensure accuracy in this podcast?” one asked.
“It’s a total disaster,” another told Status. “I think the newsroom is embarrassed.”
The paper’s head of standards Karen Pensiero wrote in an internal message to staff shared with Semafor that the situation was “frustrating for all of us.”
Readers have been noticing the errors, too. Jane Rosenzweig, a writer who covers tech, complained on Bluesky that WaPo’s AI podcaster “announced they would be discussing ‘whether or not people with intellectual disabilities should be executed’ without mentioning any context until later.”
According to Semafor, there’s a significant disconnect between the newsroom and the Post’s product division. The podcast’s product team sees the errors as a normal part part of rolling out a new and still experimental feature. The journalists, evidently, see it as an insult to their very profession.
The podcasts are developed in collaboration with the AI voice cloning company Eleven Labs, and represent the latest way that the newspaper has incorporated AI tech under Jeff Bezos’ ownership. The Post has already been using AI to provide summaries of its stories, and put forth a plan for letting non-professional writers submit articles written with AI. It also has a dedicated “Ask The Post AI” page for fielding questions to a chatbot trained on its articles.
More on AI: Time Magazine Deploys AI “Ask Me Anything” Box That Covers Up Its Actual Journalism and Can’t Be Closed
