Stories

Facebook changing deepfake policies

Illustration: Sarah Grillo/Axios

Facebook is tightening its policies on "manipulated media," including deepfakes, Monika Bickert, the company's vice president of global policy management, says in a blog post.

Why it matters: Facebook has been criticized for the way it enforces its policies on deepfakes (AI-generated audio and video) and other misleading media. In particular, critics took aim at the tech giant's decision to allow a doctored video of House Speaker Nancy Pelosi to remain on its platform last year.

What's new: The new standards call for manipulated videos to be removed from the platform if they meet the following criteria:

  1. If video is edited or synthesized beyond adjustments for clarity and quality in ways that are not obvious to an actual person and would likely mislead someone into thinking that a subject of the video said something that they did not actually say.
  2. If the video is altered via artificial intelligence or machine learning, not just Photoshop or another standard video-editing program, in a way that makes it appear to be authentic.

Between the lines: The policy updates don't extend to content that is parody or satire, Facebook says. Nor does it extend to video that's been edited "solely to omit or change the order of words."

Yes, but: Even if a video doesn't meet this new criteria for removal, it can still be taken down if it violates any of Facebook's other Community Standards covering issues like graphic violence, nudity or hate speech.

  • Similarly, any type of media that is identified as being a part of a coordinated inauthentic campaign will be taken down — even if the video doesn't violate the deepfake policy.

What's next: Videos that don’t meet the new standards for removal are still eligible for review by Facebook's third-party fact-checkers, the company says.

The bottom line: Facebook says that it doesn't want to remove all manipulated videos flagged by fact-checkers as false because those videos will be available elsewhere on the internet regardless. Rather, it thinks the better policy is to leave them up and label those videos as false — giving users context that they may not get elsewhere.