Bulletins

Twitter has a new set of deepfake rules.

Months after the company published draft guidelines for dealing with what it calls "synthetic and manipulated media," it's now finalized them. On March 5, it will start enforcing them.


  • Twitter will consider three criteria for media on its platform: Is it wholly fake or manipulated? Is it being shared in a deceptive manner? And is the media likely to affect public safety or cause serious harm?
  • An image or video or whatever else — Twitter executives were careful to say that these policies are not about one type of media or one technology for fakery — that passes all three tests will probably be taken down. Fake content of lesser consequence will be left up, but will have a label attached saying it's been manipulated or fabricated.
  • What counts as "synthetic and manipulated"? Yoel Roth, Twitter's head of site integrity, offered a few examples on a call with reporters: videos that are selectively cropped or edited; videos where subtitles are changed or removed; videos that are slowed down to change their meaning; images shared in the wrong context.
  • Twitter would have left up that slowed-down Nancy Pelosi video from last year, for instance, but it would have attached a label warning that it had been changed. But it might have taken down specific tweets with the video, depending on their context.
  • Roth also said that Twitter is trying to be more proactive in taking these actions before people see and share content, but admitted that Twitter still largely relies on users to both find and fact-check content.
Latest Stories