Twitter applied its new “manipulated media” label for the first time on Sunday to a deceptively edited video of former vice president Joe Biden. The video was shared by White House social media director Dan Scavino and retweeted by President Donald Trump.

The video was the first test of a new policy the social media company implemented on Thursday to label tweets that contain manipulated or synthetic media, ranging from edited videos to more sophisticated examples known as “deepfakes” that can fabricate events that never happened.

In this case, the altered video of Biden – who has surged to the front of the Democratic race to oust Trump in November – is based on a speech he gave in Kansas City, Missouri, on Saturday. It was then shared on Twitter by Scavino, only edited to make it appear as if Biden inadvertently endorsed Trump for reelection.

The version of the video shared by Scavino showed Biden stumbling on a line during a speech, then saying, “Excuse me. We can only reelect Donald Trump.”

But the edited video deleted the second part of the former vice president’s sentence. The whole thing said: “Excuse me. We can only reelect Donald Trump if in fact we get engaged in this circular firing squad here. It’s gotta be a positive campaign.”

Twitter applied the label to Scavino’s tweet at about 5 p.m. on Sunday evening, about 18 hours after Scavino first shared the video. The video had at least 5 million views and more than 21,000 retweets as of Sunday evening.

Twitter’s rollout of the new label was not without technical glitches, however. The label was not showing up when people searched for Scavino’s tweet, though Twitter spokeswoman Katie Rosborough said it was appearing in individuals’ timelines. She added that the company is working on a fix.

The “misleading” label is one way social media companies are trying to crack down on false and misleading information in 2020, following a 2016 election in which they were widely blamed for allowing incorrect information to widely circulate on their platforms, influencing the election and providing Russian trolls and bots with entry into the American political system.

But the companies are not acting in tandem and their policies are inconsistent. Facebook, too, has a policy for manipulated video, and it says it will remove fake videos that meet certain criteria. It also has partnerships with third-party fact checkers, and applies labels to videos those organizations determine are false.

Twitter’s new policy prohibits sharing synthetic or manipulated media that could cause harm. But like in this instance, the company may apply labels to tweets to help people understand their authenticity or to provide additional context.

It’s rare for Twitter to take action against tweets shared by Trump, though there have been complaints that the president’s tweets violate the company’s policy. Twitter has previously taken action against the president’s tweets for copyright violations.

The White House and the Biden campaign did not immediately respond to requests for comment.