Facebook announces ban on deepfakes, but it won’t stop misinformation

Facebook has announced a new policy designed to deal with misinformation spread by deepfakes. However, the website’s fairly weak rules still fall short of dealing with misinformation itself.

In a blog post published on Monday evening, Facebook’s vice president for global policy management Monika Bickert stated that the company is “strengthening [its] policy toward misleading manipulated videos,” including deepfakes as well as “all types of manipulated media.”

“Going forward, we will remove misleading manipulated media if it meets the following criteria,” wrote Bickert.

“It has been edited or synthesized — beyond adjustments for clarity or quality — in ways that aren’t apparent to an average person and would likely mislead someone into thinking that a subject of the video said words that they did not actually say.

“And: It is the product of artificial intelligence or machine learning that merges, replaces or superimposes content onto a video, making it appear to be authentic.”

Bickert further notes that parody and satire are excluded from these new rules, as is “video that has been edited solely to omit or change the order of words.”

Though this sounds like a decent first step, Facebook’s new policy still leaves the door wide open for misleading videos made through less advanced video manipulation methods. And so far, these clips have had a far more significant impact on U.S. political discourse.

In Nov. 2018, the White House shared a video of CNN reporter Jim Acosta seeming to chop his hand down on a White House intern’s arm, using the footage to justify barring him from the grounds. However, White House counselor Kellyanne Conway later admitted that the footage had been sped up, which made Acosta appear more aggressive than he was. Acosta also said “pardon me, ma’am” to the intern during the exchange, which was removed from the altered footage.

Despite the clip of Acosta being a misleading manipulated video, Facebook’s new rules would still allow the footage to continue circulating simply because there was no machine learning involved in its creation. 

Other misleading videos that would still be permitted under Facebook’s policy include footage of Speaker Nancy Pelosi that was slowed down to make her appear drunk, and selectively edited video of presidential candidate Joe Biden that made him appear to make a racist comment.

Bickert stated that media which doesn’t meet the criteria for removal may still be flagged for review by a fact-checker. If found to be at least partially false, Facebook may then “significantly reduce” circulation of the video or image, as well as apply a warning to it.

“If we simply removed all manipulated videos flagged by fact-checkers as false, the videos would still be available elsewhere on the internet or social media ecosystem,” wrote Bickert. “By leaving them up and labelling them as false, we’re providing people with important information and context.”

However, the same could be said of deepfakes. Considering the damage misleading footage can do, Facebook’s refusal to remove media unless it is both misleading and the result of artificial intelligence seems like an arbitrary distinction. As digital forensics expert Hany Farid told the Washington Post: “Why focus only on deepfakes and not the broader issue of intentionally misleading videos?”

It’s important that we deal with misleading deepfakes as the technology becomes more prevalent, particularly as the U.S. heads toward a presidential election. However, it’s no less crucial to stop less technologically advanced deceptions as well.