Meta's Oversight Board calls failure to label manipulated Iraq content 'unjustifiable'

Meta's Oversight Board has criticised the Facebook parent company for its “unjustifiable” failure to label an AI-manipulated audio clip purporting to portray two Iraqi Kurdish politicians discussing election rigging.

Created in 2018 to provide independent oversight on Meta's content moderation and how the company treats its customers, the board said it had decided to require that Meta label the audio clip as being manipulated by AI, adding that the tech conglomerate, which also owns Instagram and WhatsApp, was wrong for failing to do so.

“Meta’s failure to automatically apply a label to all instances of the same manipulated media is incoherent and unjustifiable,” the Oversight Board wrote in a report outlining its decision.

The board further found Meta should make labels for manipulated media available in local languages already available on its platforms.

“This should, at the least, form part of Meta’s electoral integrity efforts,” the board continued, also saying the company hadn't yet made available its manipulated media label in the Sorani Kurdish language.

The post in question came two weeks before Iraqi Kurdistan elections in October 2024 and garnered more than 200,000 views.

Some viewers flagged the audio clip to Meta as misinformation based on it being manipulated. But the company, citing an internal “classifier score”, decided not to flag the audio as AI-generated.

This came several months after Meta announced it would start labelling AI-generated audio, image and video content as “Made with AI” to address the issue of misleading content on its platforms.

Without detailing how it would work, the company however said it would only label AI-generated content – and would not remove it – unless it violated its policies or in the “highest risk scenarios”.

Regardless, for reasons not made entirely clear, the audio purporting to show two Iraqi Kurdish politicians was never given an AI label.

“Meta’s failure to deploy the tools it has to automatically apply the 'AI Info' label to all instances of the same manipulated media is incoherent and unjustifiable,” the Oversight Board report read.

But Meta's Oversight Board has lost considerable clout over the past year.

Meta's decision in January to end its third-party, independent fact-checking programme caught the board by surprise, and was considered by some to render the once influential panel something of a lame duck.

The board was blunt in its response to those changes.

“Meta's policy and enforcement changes were announced hastily, in a departure from regular procedure, with no public information shared as to what, if any, prior human rights due diligence the company performed,” it said.

Meta has also announced it will not be acting on an Oversight Board decision related to two pieces of content that some considered to fall into the category of hate speech, bullying and harassment.

“The posts should nonetheless be allowed upon escalation in our content review process, given their newsworthiness,” Meta said, seemingly confirming speculation that the board has lost much of its influence.