CNN
—
False claims that President Joe Biden fell asleep during a moment of silence for the victims of the Maui wildfires. Conspiracy theories that the recent spike in COVID-19 cases is being engineered by Democrats ahead of the election. Obituaries for deceased NBA players calling them “useless.”
These bogus, bizarre stories aren't published in some far-flung corner of the Internet. They're published by Microsoft. The company's homepage, also known as MSN.com or Microsoft Start, remains one of the most visited websites in the world and where millions of Americans get their daily news.
But Microsoft's decision to increasingly rely on automation and artificial intelligence over human editors to edit its homepage appears to be the cause of a recent surge in false and bizarre articles on the site, a person familiar with how the site works told CNN.
The site is preloaded as the default start page on devices running Microsoft software, including Microsoft's latest “Edge” browser, the successor to the company's Internet Explorer browser, and in 2018 employed more than 800 editors to select and edit news stories for publication to millions of readers around the world.
But in recent years, Microsoft has been laying off editors and telling them some will be replaced by “automation,” understood to be AI.
Microsoft's early investment in OpenAI, maker of the popular ChatGPT app, puts the company at the forefront of a promising, yet potentially dangerous, AI revolution, as the tech giant's president speaks publicly about responsible use of the technology.
But the role that Microsoft's AI played in recent fake news stories has raised questions about the company's public embrace of the emerging technology, and the journalism industry as a whole.
Concerns and tensions over Microsoft's use of AI in news content came to a head on Tuesday when the company accused The Guardian of damaging its reputation by publishing an article on its site.
To power its highly-visited news portal, Microsoft has signed licensing deals with major news organisations around the world, including The Guardian and CNN, allowing it to republish their articles in exchange for a cut of advertising revenue.
Last week, The Guardian published an article about Lily James, a 21-year-old woman who was found dead with severe head injuries in her school in Sydney, Australia.
Ms James' murder caused widespread grief and sparked a nationwide debate about violence against women in Australia.
But when MSN republished the Guardian article, it accompanied it with an AI-generated survey asking readers: “What do you think was the cause of the woman's death?”, giving three options: murder, accident or suicide.
The poll drew criticism from Microsoft readers: “This is the most pathetic and disgusting poll I have ever seen,” one reader wrote.
In a scathing letter to Microsoft and seen by CNN, Guardian Media Group chief executive Anna Bateson said auto-generated polls were a “deeply concerning” application of the company's AI technology and demanded that the company take full responsibility.
“An application of this nature not only has the potential to cause distress to the families of the individuals subject of the stories, but it also causes serious damage to The Guardian's hard-earned reputation for trustworthy and sensitive journalism, and to the reputations of the individual journalists who wrote the original stories,” Bateson said in a letter to Microsoft President Brad Smith on Tuesday.
Bateson said Microsoft's use of AI-generated polls alongside articles was “exactly the kind of thing we've been warning about in news” and “a key reason why we previously asked your team not to apply Microsoft's experimental genAI technology to journalism licensed from the Guardian.”
A Microsoft spokesperson said the company had disabled all polls on news articles and that it was “investigating the cause of the inappropriate content. Surveys should not appear on these types of articles. We are taking steps to prevent this error from happening again in the future.”
But the uncomfortable poll wasn't the first public misstep resulting from Microsoft's embrace of AI.
In August, MSN published a false claim on its homepage that President Joe Biden had fallen asleep during a moment of silence for victims of the devastating Maui wildfires.
The following month, Microsoft republished an article about former NBA player Brandon Hunter, who had died suddenly at age 42, with the headline “Brandon Hunter: Useless at 42.”
Then, in October, Microsoft republished an article claiming that San Francisco Supervisor Dean Preston had resigned following criticism from Elon Musk.
The story was a complete lie.
Some of the articles Microsoft featured were originally published by unknown websites and may have gone unnoticed amid the deluge of online misinformation that circulates daily.
But Microsoft's decision to republish articles from non-mainstream media exposed them to potentially millions of additional readers, breathing life into their claims. A former editor who worked at Microsoft told CNN that these types of false articles, and nearly all of them from low-quality websites, would never have been featured by Microsoft if the company hadn't used AI.
Lynn Pfeuffer, who had worked as a contractor for Microsoft on and off for eight years, said she received a call in May 2020 informing her that her entire team was being laid off.
A Microsoft spokesperson told CNN in a statement on Wednesday that 2020 was the year the company began its transition to a “personalized feed” that is “algorithmically tailored to viewer interests.”
As the site's lifestyle editor, Pfeuffer's job involved selecting and curating stories to feature across Microsoft's sites.
“Our editorial team is really tight-knit, incredibly talented and has been working together for a long time,” Pfeuffer told CNN. The group, which included veteran journalists, took their curatorial responsibilities seriously, she said.
“How many people [MSN]”We had to be responsible with what we put on the site because a lot of people could read it and be affected by it,” she told CNN.
Microsoft declined to say how many human editors still curate the site, or whether it's all done by AI and algorithms.
Scrolling through the Microsoft homepage a few weeks ago, Pfeuffer said it looked different from the portal she and her colleagues once managed: Not only were objectively false and outlandish articles posted on the site, but it also featured overtly partisan articles with headlines like “Should America Get Rid of Biden Forever?” and “The Dark Side of the Unfolding Disaster Under Joe Biden.”
The articles CNN reviewed on Microsoft's homepage were not published by reputable, established news organizations, but rather came from smaller sites that provided little or no information about who wrote the articles or details about their editorial standards.
Pfeffer said these sites and headlines were never prominently featured on the company's homepage while she and her editorial colleagues worked at Microsoft, and she was horrified to see “these websites that have the appearance of fact, but are not legitimate news sources, putting misinformation on our homepage.”
“We tried to cover everything fairly, from all sides, and not take a political stance on the stories,” she told CNN.
It was sensationalist headlines like these that caught the attention of Ferris Kawar, a father of two and sustainability manager from Santa Monica, California.
After a recent software update, Kawar's default web browser has changed to Microsoft Edge and with it, the homepage has also switched to Microsoft.
“It was like standing in line at the grocery store reading the front page of the National Enquirer,” Kawal told CNN.
“I think Microsoft is a pretty trustworthy company,” he said, “and I don't think of them as political at all. But the content being pushed here seemed clearly right-leaning to me.”
Microsoft and its president, Brad Smith, have published extensively about the responsible use of AI.
But the company provided few specifics when asked by CNN.
“As with all our products and services, we continue to adjust our processes and are constantly updating existing policies and defining new ones to keep up with emerging trends. We are committed to addressing recent issues with low-quality articles posted to Feed and are working closely with our content partners to identify and address issues and ensure they meet our standards,” a Microsoft spokesperson said in a statement on Wednesday.
But in a letter sent to the company on Tuesday, the Guardian said the company had “failed to substantively respond” to the paper's questions about how AI would be used in journalism.
Microsoft is no longer Kawal's homepage, and he has changed his default browser back to his preferred Google Chrome, but millions of people around the world still get their news from Microsoft.
“If newspapers have a news-only front page, they should take that responsibility seriously and understand that it has an impact on society,” Kawal said.
