Oversight committee calls Meta’s unbalanced AI temperance “inconsistent and unreasonable”
With Meta’s platform full of more AI-generated content, the company still has a lot of work to do when it comes to enforcing its policies around manipulating the media. The Oversight Committee has again criticized social media companies for handling such posts and wrote in its latest decision that it is “inconsistent and unreasonable” to be unable to enforce its rules at all times.
If it sounds familiar, it’s because this is the oversight committee since last year using the term “inconsistency” to describe the way Meta manipulates the media. The board had previously urged Meta to circulate on Facebook on a misleading edited video of Joe Biden to update its rules. In response, Meta said it uses tags to identify content generated by AI and will apply more prominent tags in “high risk” situations. These tags (such as those below) are noted when using AI to create or edit posts.
The board said this approach is still lacking. “The board of directors is concerned that despite the increasing popularity of manipulation across formats, Meta’s implementation of manipulation policies is inconsistent,” it said in its latest decision. “It is incoherent and unreasonable for Meta to fail to automatically apply labels to all instances of the same manipulation media.”
The statement puts forward a decision related to an article that was audio from two politicians in Iraqi Kurdistan. The so-called “recorded dialogue” includes discussions about manipulating upcoming elections and other “sincere plans” in the region. The board said the post had reported misinformation to META, but the company “no human review” would end the case. Meta later tagged certain instances of audio clips, but were not originally reported.
According to the board, the case is not an outlier. Meta obviously told the board that it does not automatically recognize and apply tags to audio and video posts, but only for “static images”. This means that multiple instances of the same audio or video clip may not get the same processing, and the board notices that this may cause further confusion. The Oversight Committee also criticized Mehta for often relying on third parties to identify AI-Warmed video and audio, just as in this case.
“Given that META is one of the world’s leading technology and AI companies with its resources and extensive use of the Meta platform, the board reiterated that Meta should prioritize investment in technology to identify and mark video and audio manipulation on scale,” the board wrote. “It is unclear why this technical expertise and resource outsourcing company will be identified in high-risk situations that may be manipulated by media or trusted partners.”
In its recommendation to META, the board said the company should consistently mark the “clear process” of “same or similar content” in the case when adding a “high risk” tag to posts. The board also recommends that the tags should appear in languages that match the rest of the settings on Facebook, Instagram and Threads.
Yuan did not respond to a request for comment. The company has 60 days to respond to the board’s recommendations.
