Meta's auditing agency deems the video of Joe Biden reaching for his granddaughter's chest as ``safe because it is not made by AI and does not violate policy,'' but complains that the current incoherent policy should be improved.
Meta's Oversight Committee has ruled that a video edited to appear to show President Joe Biden inappropriately touching his granddaughter's breasts violates Meta's policies. The decision was made that 'it is not.'
Oversight Board Upholds Meta's Decision in Altered Video of President Biden Case | Oversight Board
Oversight Board calls on Meta to rewrite 'incoherent' rules against faked videos | TechCrunch
https://techcrunch.com/2024/02/05/meta-facebook-oversight-board-biden-video-cheapfake/
The video in question was an edited version of footage of President Biden taken during the midterm elections in October 2022. At the time, President Biden celebrated his adult granddaughter's first vote by placing an 'I Voted' sticker on her chest and kissing her on the cheek, following her instructions.
However, the video was altered to show President Biden touching her breasts on a loop, and while a song with the lyrics 'Girls, play with your breasts' was played in the background, the video was edited to show President Biden touching her breasts, and a song with the lyrics 'Girls, play with your breasts' was played in the background. The video was clearly edited in a way that belittled the president, with subtitles saying, 'The people who voted are mentally ill.'
Another user reported this post to Meta as hate speech, and Meta automatically processed the report without any review. The user appealed to Meta again, dissatisfied with this decision, but a human reviewer determined that the content was not in violation, and the post was left alone. Eventually, the users complained to the Oversight Committee.
Facebook's independent audit agency begins discussion of policy review after Meta refuses to delete video labeling President Biden as a ``pedophile'' - GIGAZINE
The Oversight Board is an independent adjudication body established by Meta (then Facebook) in 2020.
The Surveillance Committee continued its investigation for a while, seeking public comments, but ultimately decided that ``There was no problem with Meta leaving the video in question.''
According to the monitoring committee, Meta's information manipulation policy at the time of writing only prohibits ``videos that have been edited in a way that makes it appear that the person said something they did not say.'' , furthermore this only applies to 'videos created by AI'. According to Meta, the main characteristic of 'manipulated media' that the policy defines as a violation is that it 'may mislead the average user into believing that it is authentic and unaltered.' The company will only take action based on its policy if such misleading content is produced by AI.
This video is ``a video that makes it look like someone did something that no one has ever done,'' and in addition to being created by a human rather than an AI, one scene of the video is looped. It is clear that it has been modified. Therefore, the Surveillance Committee has determined that the video in question is ``no problem.''
However, the Oversight Committee is concerned about the current policy on information manipulation, stating that ``the current policy is disjointed and lacks convincing justification, rather than being aimed at preventing specific harm. , the focus is on how the content was created, which is inappropriate.'
The Oversight Committee said, ``As it currently stands, it is too narrow to apply the policy only to content that has been altered or generated by AI, and to content that appears to be saying words that a person did not say. The policy should also be extended to content that appears to be doing something that is not intended to be the case.''Based on the knowledge gained through public comments, he added, ``Altered content that does not rely on AI is widespread. Policies should not differentiate between AI-generated 'deepfakes' and non-AI-generated content.'' 'The current policy does not clearly identify the harm caused by misinformation and needs to be revised as soon as possible.'
As specific measures, the Oversight Committee stated, ``Anything that interferes with voting rights or the execution of public duties should be collectively defined as ``manipulated media'''' and ``If there is no policy violation, the content will be deleted.'' Instead of doing so, they suggested adding a label stating that it may cause misunderstandings. The Oversight Committee opined that 'Given the upcoming elections in 2024, Meta should urgently reconsider this policy.'
Related Posts:
in Web Service, Posted by log1p_kr