Twitter's Content Filters Flagged 'Union Buster' as 'Offensive Language'

It was a mistake, the company said, but highlights the imprecision of automated language filters.
Image: Anna Moneymaker-Pool/Getty Images
On the Clock is Motherboard's reporting on the organized labor movement, gig work, automation, and the future of work.

On Wednesday, Twitter marked a tweet calling Office of Management and Budget head nominee Neera Tanden a "union buster" as containing "potentially harmful or offensive language." Twitter told Motherboard that the move, which was part of its rollout of an experimental feature to curb toxic behavior on the platform, was a mistake.


In response to a tweet from Minnesota Attorney General Keith Ellison, one Twitter user tried to add "She's a union buster, Keith." Twitter's experimental system then asked the user if they "Want to review this before Tweeting?"

"We're asking people to review replies with potentially harmful or offensive language," the pop-up adds.

In a statement, a Twitter spokesperson told Motherboard that "we're still in the early testing phase and this was a mistake. People who think they've received a prompt by mistake can share their feedback with us so we can improve."

In 2019, the Center for American Progress, of which Tanden is president, laid off the entire union staff of its ThinkProgress newsroom. At the time, CAP said it intended to continue operating the site without union labor.  

Twitter announced the relaunch of the experimental feature earlier this month.