Meta Oversight Board orders removal of deepfake videos from Facebook
What's the story
Meta's Oversight Board has directed the tech giant to take down deepfake videos of a UK politician and a young Muslim woman from Facebook. The board found that the Scotland councilor video violated Meta's rules on hateful conduct by falsely depicting criminal and predatory sexual behavior by refugees as an entire group, not individuals. It also said that Meta's current safeguards are "consistently and fundamentally inadequate" to deal with the rising tide of AI deepfakes.
Criticism
Board criticized Meta for not labeling video as AI-generated
The board criticized Meta for its decision not to label the video of the Scottish politician as AI-generated.
The video showed the politician making inflammatory comments about refugees, which was later found to be a deepfake.
Despite being flagged directly by the board, Meta maintained that it didn't violate its content policies and didn't warrant an AI label.
Recommendations
Board calls for stricter measures against deepfakes
The Oversight Board has called for stricter measures against deepfakes, including expanding the criteria for applying 'high risk' labels.
It also suggested more action to curb the spread of misleading AI content, harsher penalties for accounts that repeatedly share it, and greater transparency on data regarding when AI labels are applied.
The board proposed nine policy changes at Meta in total.
Policy changes
Recommended changes to ensure algorithms demote 'high risk' content
Among the proposed policy changes, the board suggested that Meta should ensure its algorithms demote content labeled "high risk." This would make such posts less likely to appear in users' feeds.
The board also recommended making it harder for users to view AI-generated content by introducing a warning screen that requires a click-through before viewing the content.
Another case
Separate case involving deepfake video of Muslim woman
In a separate case, the board also ordered Meta to take down an AI-generated video mocking a young Muslim woman in Europe.
The woman was participating in a campaign to improve menstrual health education and reduce stigma around health conditions for women and girls from ethnic minority backgrounds.
The board recommended changing its definition of "unwanted manipulated imagery" to include deepfakes of private individuals saying or doing things they did not say or do.