Microsoft’s AI news tool is causing chaos

Last week, we reported on an incident with Microsoft’s MSN news aggregation platform that’s now partly powered by artificial intelligence (AI). Beside an article from The Guardian newspaper that was shown on the MSN site, the AI tool autogenerated a poll. The article was about a woman who was found dead in a school bathroom, and the poll asked readers what they think caused her death: murder, accident or suicide. After this The Guardian said it was damaging to its reputation (and also hurtful to the deceased’s family). Microsoft has since suspended Microsoft-generated polls for all news articles on MSN and said that it’s “investigating the cause of the inappropriate content.”

But this poll isn’t the only problem with Microsoft’s MSN news AI. Apparently, Microsoft has been cutting down its human news editors over the past years, and some were even directly told that they’d be replaced by “automation” ( likely referring to AI). 

“We had a really tight knit, super talented editorial team, and we had all worked together for a long time,” said Ryn Pfeuffer, who worked as a contractor for Microsoft for eight years and whose entire team was laid off in May 2020, to CNN. “I don’t think people realize how many people use [MSN],” she added. “You had to be responsible what you put on the site because so many people would read it and could be swayed by it.”

According to a new CNN report, MSN’s AI is literally making up fake or misleading news. It featured a story that falsely claimed US President Joe Biden fell asleep during a moment of silence for victims of Maui’s wildfire, republished one about deceased NBA player Brandon Hunter with the headline “Brandon Hunter useless at 42” and republished another claiming that San Francisco Supervisor Dean Preston had left his position after being criticized by Elon Musk – which was totally untrue. That’s not to mention obviously fabricated stories like ones about fishermen catching mermaids and Bigfoot sightings. It’s also had problems with getting stories from unreliable sources. And, it’s published AI-generated content, like a travel guide to Ottawa, Canada, that suggested tourists visit a food bank "on an empty stomach." Soon after posting this, Microsoft had to backtrack. 

MSN isn’t the only news platform experimenting with AI. The BBC has started developing its own AI tools for its site, and even The Associated Press has used AI to build its “Automated Insights” for almost a decade at this point.

Microsoft has some major investments in OpenAI (the firm behind ChatGPT and Dall-E), so the move toward adopting this kind of tech into many of its products makes sense. But it still begs the question, would human editors allow these mistakes to reach news consumers?