/cdn.vox-cdn.com/uploads/chorus_image/image/51888597/REC_ASA_CODE2016-20160601-142517-2447.0.0.jpeg)
Twitter CEO Jack Dorsey apologized Thursday morning after an ad promoting white supremacy appeared in users’ timelines Tuesday evening.
“We made a mistake here and we apologize,” Dorsey tweeted. “Our automated system allowed an ad promoting hate. Against our policy. We did a retro and fixed!”
We made a mistake here and we apologize. Our automated system allowed an ad promoting hate. Against our policy. We did a retro and fixed! https://t.co/7gvycmzpsm
— jack (@jack) November 17, 2016
The ad was first reported by Twitter user Ariana Lenarsky:
.@twitter I can't believe anything still surprises me, but why the fuck am I seeing nazi ads on this website pic.twitter.com/jtKKnn8XR5
— Ariana Lenarsky (@aardvarsk) November 16, 2016
It has since been removed, and the user who posted it has been suspended.
But it’s not a great look for Twitter, which earlier this week rolled out new features to try and curb abuse on its platform. The company was also apparently in disbelief that the ad even ran, suggesting to BuzzFeed News that it might be either old or photoshopped. It was neither.
@jack @motherboard pic.twitter.com/e8c8XTFurF
— John Paczkowski (@JohnPaczkowski) November 17, 2016
The author of that BuzzFeed News story was not impressed with Dorsey’s apolotweet.
thank you for fixing, @jack. but why did your spokesman tell me “it looks like the screenshot in that tweet is either old or photoshopped”?? https://t.co/DZv3jnVLRn
— Charlie Warzel (@cwarzel) November 17, 2016
Update: A Twitter spokesperson provided a statement in addition to Dorsey’s tweet.
"Our ads policy prohibits abusive content, hateful conduct, but our automated system missed one this past Monday - we apologize. The Promoted Tweet was live for less than an hour before we removed the Tweet. We've identified where our process broke down and are making immediate changes to prevent this from happening in the future. Specifically, we're going to be adding more keywords and image recognition parameters to more quickly flag this type of content for review."
This article originally appeared on Recode.net.