Zuckerberg says it’s “really important” for Facebook to create deepfake videos policy

Facebook chief operating officer Mark Zuckerberg said the social network is evaluating how it should handle “deepfake” videos created with AI and high-tech tools to yield false however realistic clips.

In an interview at the Aspen ideas festival in Colorado on Wednesday, Zuckerberg said it might make sense to treat such videos otherwise from other misinformation like false news.
Facebook has long held that it shouldn’t decide what’s and isn’t true, leaving such calls instead to outside fact-checkers. However Zuckerberg said it’s worth asking whether or not deepfakes are a “completely different category” from regular false statements. He said developing a policy regarding these videos is “really important” as AI technology grows more refined.


The comments come as Facebook prepares to unleash details on the company’s planned oversight board, which is meant to control speech on the 2.7 billion-member social network.


Facebook, like alternative social media corporations, doesn’t have a particular policy against deepfakes, whose potential threat has emerged solely in the last few years. Company executives have said in the past that it makes sense to look at them under the broader umbrella of false or deceptive information. However Zuckerberg is signaling that this view might be changing, leaving open the probability that Facebook may ban deepfakes altogether.


Doing so, of course, could get difficult. Satire, art and political dissent could be swept up in any excessively broad ban, making a lot of headaches from Facebook.


Other false videos might still get a pass. As an example, the recent altered video of House Speaker Nancy Pelosi that made her sound like she was slurring her words doesn’t meet the definition of a deepfake.


In Adam Mosseri’s initial U.S. TV interview since taking over as head of Facebook-owned Instagram, CBS News’ Gayle King on Monday asked him about a deepfake video of Zuckerberg that seems to show him talking regarding amassing and wielding power by owning people’s information on CBS News’ streaming platform, CBSN. The real version of the video aired on CBSN in September 2017. Facebook hasn’t removed the fake video, despite CBS asserting trademark infringement and asking it to take down the video for copyright violation.


“Well, we don’t have a policy against deepfakes presently,” Mosseri said. “We are trying to evaluate if we wanted to do that and if so, how you’d outline deepfakes.” He added, “We need to have outlined principles, and we got to be clear regarding those principles.”