Abuse of Facebook's use of artificial intelligence for content moderation

Failure in Meta's use of AI for content moderation under Community Standards:

On 12 April 2024, between 12:27 am and 12:57 am CET, my profile account, Karim Nader (https://www.facebook.com/QadirNader/), received 23 notifications from Meta that 23 of the contents that I posted on Facebook went against the community standards. I appealed all the decisions by Meta, and only one of the 23 appeals was approved soon after I submitted. Below is the screenshot of the result of the appeal that I made.

Meta still needs to review the rest of the appeals. The reports I received from Meta are all within the same timeframe, which confirms a pattern of the attack against me since the contents are posted at different times, but they are all flagged by Meta at the same time. Below are the times when I received the notifications:

Friday, April 12, 2024 at 12:57 AM

Friday, April 12, 2024 at 12:52 AM

Friday, April 12, 2024 at 12:52 AM

Friday, April 12, 2024 at 12:52 AM

Friday, April 12, 2024 at 12:52 AM

Friday, April 12, 2024 at 12:47 AM

Friday, April 12, 2024 at 12:47 AM

Friday, April 12, 2024 at 12:47 AM

Friday, April 12, 2024 at 12:42 AM

Friday, April 12, 2024 at 12:42 AM

Friday, April 12, 2024 at 12:42 AM

Friday, April 12, 2024 at 12:37 AM

Friday, April 12, 2024 at 12:37 AM

Friday, April 12, 2024 at 12:37 AM

Friday, April 12, 2024 at 12:37 AM

Friday, April 12, 2024 at 12:37 AM

Friday, April 12, 2024 at 12:37 AM

Friday, April 12, 2024 at 12:37 AM

Friday, April 12, 2024 at 12:37 AM

Friday, April 12, 2024 at 12:32 AM

Friday, April 12, 2024 at 12:27 AM

Friday, April 12, 2024 at 12:27 AM

Based on the above timeframe, it appears that a third-party informed Meta that I had posted content that violated their community standards. The third party submitted all their reports at once, which is reflected in the timeline of when I received notifications from Meta that my posts went against their community standards.

Below are three examples of what Meta marked as violations of its community standards.

Failure in Meta's use of AI for Copyright Infringement and Intellectual Property:

My profile account received two notifications from Meta about copyright infringement on April 12, 2024. Both notifications were received at 21:41 PM CET, which was the same day that I received notifications for violating community standards. Although the copyright infringement notifications were received at a different time than the community standards notifications, they still support the pattern that they came from the same third party. The content that has been reported for copyright is an online interview I conducted with the deputy governor of Sulaymaniyah city in Iraq. The report number is: 44467622806813.

The original video is posted here on my YouTube: https://www.youtube.com/watch?v=57ERUP8cO_Q and please check the metadata of the video here: https://mattw.io/youtube-metadata/ . I have also posted a copy of the video on my Facebook(Karim Nader).

Below is a screenshot of my previous videos; the person on the right is me(Meta has my passport, my ID, my picture and my video) and the person on the left is the government officials that I interview remotely. Then after obtaining the consent, I publish the interviews on my social media platforms including Facebook both on my account profile and page.

My Facebook profile account and pages are full of similar videos, this shows how AI fails to assess the request by the reporter.

This is also my YouTube wall you can see all these videos have been posted and YouTube channel is verified by YouTube.

The second copyright infringement report I received was on April 12, 2024, the content is a video of myself recorded in the backyard of my house in Stockholm. The report number is 444676228068131.

Who is the fake copyright holder?

Meta has stated in its message to me that the name of the copy reporter is [email protected], and although Wllat Media is a Kurdish social media outlet that I know very well, I am certain that they are not the reporter. Instead, it seems that someone or a party is impersonating Wllat Media. To investigate further, I followed a few steps - I copied and pasted the reporter's name and searched for it on Facebook. The search result that I obtained was a Facebook page called Wllat Media, which had the same email address in its about section.

Now the reporter has been located, and there are several reasons why the page wllat media is not authentic. Based on data from meta, Wllat Media was created in 2020, but there is no information on whether the name of the page has been changed. Meta doesn't show the chain of changes made to a page if it was merged with another page, so we cannot view the previous name of the page. However, by examining the contents, videos, and images posted on the page, it is evident that the page initially posted in Bangali and later started posting in Arabic and Kurdish. The page was first posted in Arabic on August 31, 2022. However, the edit history of the first Arabic post shows that it was edited on March 7, 2024, as indicated by the screenshot.

The following bullet points provide additional details:

1. The first post to have Arabic content and a picture with Kurdish design was on August 31, 2022. However, upon checking the post's edit history, it is clear that the Arabic text and images were added on March 7, 2024.

2. The second post was on September 1, 2022, and edited on March 7, 2024.

3. The third post, from September 6, 2022, was edited on March 7, 2024, and Kurdish text and an image were added.

4. The fourth post, originally published on September 9, 2022, was edited on March 7, 2024, and Kurdish text and an image were added.

5. The fourth post, originally published on October 22, 2022, was edited on March 7, 2024, and Kurdish text and an image were added.

6. The fourth post, originally published on October 23, 2022, was edited on March 7, 2024, and Kurdish text and an image were added.

The page's last post was on March 7, 2024, and all content posted was taken from Wllat Media's original page. Below is another screenshot of how the editing was done.

By editing old posts, third party agencies manage to abuse AI and claim fake and unfair copyrights against journalists and human rights activists.

If you search for Wllat Media on Facebook, there is one page with 100K followers and then there is the page that reported copyright infringement against me. Although, I could not locate the two videos on the fake wllat media page, however, its possible that the attackers removed the video from the page right after the copyright infringement was approved by Meta.

I hope this article encourages more Human Rights Activist and Journalists who face similar issues to speak about it with the hope it will trigger Human Rights team at Meta.