As you scroll through Facebook, Instagram, or Threads, you’ve probably noticed more and more AI-generated posts. However, while this content grows, Meta isn’t keeping up with clear rules for manipulating media. Now, the company is again facing strong criticism from its own Oversight Board, which says Meta’s approach is still confusing, unreliable, and falling short of what’s needed.
Oversight Board highlights serious failings
In a decision published on June 24, the Oversight Board said Meta’s current methods for dealing with altered or AI-generated media are “incoherent and unjustifiable.” If this sounds familiar, the board used almost the exact wording last year when it told Meta to clean up its policy after a fake video of US President Joe Biden spread quickly on Facebook.
Back then, Meta promised to take action by expanding its labels that mark content made or changed with AI. It also said it would use stronger warnings for “high-risk” situations—those with the potential to cause harm or spread serious misinformation. These labels might say, “This content was digitally created or altered.”
However, the Oversight Board has now clarified that this plan isn’t working well enough.
“Despite the increasing spread of manipulated content, Meta’s enforcement is inconsistent,” the board said. “Its failure to automatically label every instance of the same altered media is incoherent and unjustifiable.”
A worrying case in Iraqi Kurdistan
The board’s statement came in a specific case involving a disturbing audio clip. It was posted on Facebook and claimed to be a secret recording of two politicians in Iraqi Kurdistan. In the clip, the speakers appeared to talk about election fraud and dangerous political plans. Many believed it was real.
A user flagged the post for misinformation, but Meta dismissed the case without letting a human reviewer check it first. Later, the company added a label to some versions of the audio—but not to the original post. That inconsistency, the board said, is part of a bigger problem.
According to Meta, it currently can’t automatically detect and label audio or video content—it can only do that with still images. Even if the same AI-edited video or sound clip is posted multiple times, not every version will be marked. This could be confusing and even dangerous for users like you, especially when the content is misleading or harmful.
Meta urged to take full responsibility
The board also pointed out that Meta often leaves the hard work of identifying fake or AI-altered media to outside partners or media groups. The company relied on third parties to flag the clip as potentially false or manipulated.
That’s not good enough, the board said.
“Meta is one of the world’s top tech and AI companies,” the board wrote. “It has the tools and the reach. So why does it depend on outside groups to do this job—especially in high-risk situations?”
The board has given Meta several clear recommendations. These include creating a proper system to apply the same warning labels to similar or identical content across all its platforms. These labels should also be shown in the same language as the user’s Facebook, Instagram, or Threads settings.
Meta hasn’t publicly responded to the decision. The company has 60 days to reply to the board’s suggestions and possibly take action.
