"Blue" war zone

After Israel began its military ground operation against Hamas in Gaza, on October 27, 2023, Egyptian doctor Ghada Sadek uploaded to her Facebook page a video clip showing an Israeli drone pursuing and then striking a young man.

Meanwhile, Israeli writer Tzvi Fishman published a post on his Facebook page saying: “There are no innocent civilians in Gaza.” Fishman quoted Rabbi Dov Lior, the chief rabbi in Hebron, as saying: “The Israeli army does not need to endanger its soldiers by searching houses one by one to make sure there aren’t wounded civilians. It should follow the example of America in World War II, when it wiped out two entire Japanese cities to end the war and save the lives of American soldiers.”

The video posted by Sadek stayed up only for a few hours before being deleted, and her account was suspended for 24 hours, allegedly for “violating Facebook’s standards.” But Fishman’s second post, saying that “there are no innocent civilians in Gaza,” was not deleted.

The author of this report - in cooperation with the Arab Center for Social Media Advancement (7amleh), which specializes in digital rights – set up a database register of 50 posts by Israeli citizens and public figures, led by Minister of National Security Itamar Ben Gvir, and pages affiliated with Israeli media institutions on the two Meta platforms - Facebook and Instagram. These posts contained calls for violence against civilians, the forced displacement of Palestinians, as well as a halt to humanitarian aid entering the Gaza Strip.

Legal experts we spoke to agreed that these posts constituted a violation of Meta’s policy of combatting hate speech, and were an encouragement to carry out actions prohibited under relevant international laws.

This contradicts Meta’s claim to be implementing equally across all countries, its policies “violent and shocking content,” in response to reports by rights group and the media accusing the American based company of showing bias favouring Israel by deleting Palestinian content that is considered peaceful.

In its response to this investigation’s findings, Meta said that it might make mistakes, given the fast-paced nature of the current war, but denied that there was deliberate bias towards one party over another.

But the problem does not stop here. Calls for the resumption of settlements construction have been made by some Facebook groups, which quickly moved from the realm of cyberspace to reality, with events organized on the borders of Gaza, calling for settlement construction revival in the strip.

Saida Mallah, a journalist from Morocco covering the war in Gaza, published regular updates about the war on her social media platforms. However, after she posted images of what the media dubbed the “Tents Holocaust” - a raid on Rafah by the Israeli Air Force on May 26, 2024, which left at least 45 Palestinians dead, mostly women and children - her Telegram account was suspended, and she had great difficulty retrieving it. Saida attributes this “technical glitch” in her account to her decision to post material showing human rights violations during the war in Gaza.

Saida Mallah’s post is one of 20 which this reporter has shown to have been deleted and/or restricted by one of the two Meta platforms. Most of these posts deal with human rights violations, carry news of the war, or share analytical pieces on Palestine.

In contrast, we collected 50 posts published by Israeli Facebook and Instagram accounts, following the attacks of October 7, 2023. These posts included violent rhetoric directed against Palestinians and Arabs in general. There were calls for the extermination of the people of Gaza, with no distinction made between civilians and combatants, and for stopping humanitarian aid entering the Strip. Nor were these limited to Gaza, as some users demanded that the Lebanese capital, Beirut, be turned into another Khan Yunis, in response to Hezbollah attacks in northern Israel. Some posts also put out misleading information and images.

We collected these posts, in cooperation with the “7amleh” centre and, in consultation with independent legal experts, analysed a sample of five of them to see how far the company was complying with its stated policy on hate speech and with relevant international agreements. These posts remained available on the Meta-owned Facebook platform up to the time when we confronted the company with the findings of our investigation.

When we sent Meta the posts, under its right-of-reply policy, the company deleted eight out of the total 50 posts “for violating Meta community standards.”

When Egyptian legal expert Yasser Saad reviewed the database of content suspected of containing hate speech, he said that most of it breached Article 20 of the International Covenant on Civil and Political Rights, which prohibits hate speech.

Legal researcher at the University of Antwerp in Belgium, and a specialist in genocide, Maha Abdallah agreed, saying this content constituted incitement to commit genocide and ethnic cleansing of Palestinians. She added that the management of Meta should meet its legal and moral responsibilities by taking decisive measures against such content.

The “7amleh” centre had already identified more than eight thousand cases of digital violations of pro-Palestinian content on social media sites. Removing or restricting access to this content constitutes about 40 percent of all recorded breaches of the guidelines.

The centre has meanwhile been keeping tabs on the increasing amount of what it describes as “violent” content, posted after October 7, 2023. This reached a peak on October 9, with more than 120,000 anti-Palestinian posts on social media.

Jalal Abukhater, Advocacy Manager at “7amleh”, says that permitting the posting of content that promotes hate speech and incitement in this way, poses a major threat to the lives and safety of Palestinians. He adds meanwhile that suppressing Palestinian content silences voices seeking to reveal human rights violations during the war in Gaza.

Analysing the date of publications of the posts that are the focus of our investigation shows that the majority were published in the three months following the outbreak of the war. This was followed by a dip in the intensity of postings, though those containing violent speech continued to appear at varying intervals up to June 2024.

Network analysis by this reporter of the accounts of individuals posting violent content shows that they are all followers of the account of Israeli Minister of National Security Itamar Ben Gvir, whose official Facebook page represents a central node around which the accounts in our sample revolve.

Itamar Ben Gvir plays a key influential role in this social network, which reflects more of a vibe of a homogeneous political mindset rather than an organised working group, and which has significant impact in digital interaction. Ben Gvir’s account shows a recurring pattern of inflammatory and inciteful content. Its call for Israeli settlement erection, forced Palestinian displacement, prompts strong responses and interactions from both supporters and opponents. This major interaction with his posts help to spread Ben Gvir’s content across Facebook.

In a short video - published in May 2024, on his Facebook page - the Israeli Minister of National Security called for a violent attack on Rafah, forcible displacement of the residents of Gaza, and the return of Israeli settlements construction in the strip.

Gaza City is part of the Palestinian Territories. Settlement activity there is consequently prohibited under international law, and would breach Security Council Resolution No. 2334, which calls for an end to Israeli settlements in the occupied Palestinian territories.

Despite this, Facebook did not delete Ben Gvir’s post. Furthermore the video received wide interaction, with more than a thousand followers of the page liking it, and more than five hundred accounts commenting on it.

In order to test for himself the platform’s policy regarding pro-Palestinian content and how this compares to anti-Palestinian incitement, this reporter created a Facebook account and began posting content supportive of Hamas. But this was deleted, based on Facebook’s algorithms, “for violating the Meta community’s standards governing the publishing of content supportive of groups classified as dangerous.”

He then reorientated the account and published Hebrew and English posts calling for the “destruction of Gaza.” These posts escaped the clutches of the algorithms, however, and remained present and available on the platform.

The reporter decided to press on with testing Meta’s policy and filed a report – using his personal Facebook account – complaining about the post inciting violence towards the people of Gaza. But Meta’s response came back quickly, rejecting his request, because the post “did not go against its standards.” The company asked him to adjust his account settings so as not to be exposed to such content. Facebook answered favourably to the reporter‘s request, only to remove pro-Hamas content.

The legal researcher specialised in genocide, Maha Abdallah, expressed shock when she was shown Facebook’s rejection to remove the post, that the journalist reported to Meta. The post clearly included several violations, like incitement to commit genocide, and ethnic cleansing against Palestinians, which is forbidden according to the fourth Geneva convention, and is considered incitement to commit crimes against humanity according to Rome Statute of the International Criminal Court.

We turned to Issam Hassan, a researcher in the field of artificial intelligence (AI) at the Technical University of Munich, to better understand the working of algorithms that determine whether a post should be deleted or not. He said that human intervention via algorithms comes in stages.

First, a data sample is selected for initial training. This sample may be predetermined to block specific content containing keywords. In the subsequent stages, the platform’s content goes through a screening process by oversight boards or independent organisations to identify anti-Semitic or violent content.

AI researcher Issam Hassan adds that the algorithm is retrained at regular intervals using selected content, in a process known as “algorithmic fine tuning.” The job of marginalizing, blocking and then deleting suspect content from the platform is then done automatically by the algorithm.

Hassan believes that the algorithmic model used on the two Meta platforms has been programmed to identify Palestinian content and rank it lower on the “algorithmic scoring” system. This reduces its reach and spread and even goes so far as to delete it altogether. In contrast, machine learning models can be trained on data that does not include blocking mechanisms, making it tolerant to other content. This is what may be happening with inciteful content against Palestinians, which has not been deleted by Meta algorithms.

Meta's approach to pro-Palestine content has long sparked controversy regarding human rights. Human Rights Watch (HRW) put out a report in October 2021, showing that Facebook management was increasingly deleting and applying excessive censorship to Palestinian content, as well as limiting its availability.

Facebook took seriously the findings of this report and commissioned an independent body to examine its policy on Palestinian content. The results of this review were largely consistent with those of HRW.

The review showed that “the company’s policies are having a negative impact on freedom of expression of Palestinian users, their freedom of assembly, and political participation and therefore on the ability of Palestinians to exchange information and ideas about their experiences.”

In response to these findings, Facebook promised to address its policy of censoring Palestinian content.

After the outbreak of war in Gaza on October 7, 2023, HRW put out a report showing that the same thing was happening. It documented 1,049 cases where pro-Palestinian or neutral content was being curtailed during the war, either through posts being deleted, accounts suspended, or participation and interaction on Facebook and Instagram being restricted. Meanwhile only one piece of content supportive of Israel was restricted.

The report showed that deleted posts contained no obvious violations, and in some cases carried either impartial news, calls for humanitarian relief, or expressions of personal opinion, not incitement.

It also indicated that Meta had received requests from the “Israeli National Cyber Directorate” to remove content which “Israel describes as inappropriate.” The management of Meta had responded to the majority of these requests by removing or restricting the content.

The HRW report came days after Meta announced that it was addressing errors that had limited the reach of posts in general on Instagram. The company denied having any intent to suppress “freedom of expression,” emphasising that it applied its policies on violent and shocking content equally across the world.

There have since been a series of human right investigations and press reports, documenting the continued restriction of pro-Palestine content on the company’s platforms. To make matters worse, this reporter found that the platform has continued to be used as a base to make calls for Israeli settlement activity, which quickly translated into movements on the ground.

As soon as the latest war broke out, in 2023, settlement construction calls grew more intense, parallel to Israeli settlers increased attacks on Palestinians living in the occupied West Bank.

The Israeli organisation Yesh Din, which keeps track of settlement erection calls, found evidence that groups on social media were calling for settlements construction from mid-October 2023, demanding that Palestinians “leave their lands and go to Jordan if they want to avoid a second Nakba.”

Settlers then launched a semi-organised attack on Palestinians in the town of Deir Istiya, on October 26, 2023.

Coinciding with these settler attacks, a group was launched on Facebook called “The Israeli People Are Returning to the Gaza Strip.” The introductory paragraph about the group states that it was set up “to send a clear message to our leadership, that the people of Israel demand to return and settle all of the Gaza Strip!”

The over eight thousand members of this group have regularly posted calls for events and activities aimed at completely occupying Gaza, and turning it into an Israeli settlement. The group has not allied itself to any political faction, but its administrator has regularly posted quotes from and video clips of politician Moshe Feiglin.

Feiglin was arrested during the 1990s, after organising demonstrations against the Oslo Accords, which called for a permanent settlement between Israel and the Palestinian Authority, based on UN Resolutions 242 and 338. Feiglin was subsequently given a jail sentence, but this was later commuted.

After the outbreak of the war in Gaza, Feiglin gave a television interview in which he called for the re-occupation of the Strip, echoing the narrative of the Facebook group that Israel had the right to all Palestinian territory and that settlements should be rebuilt in Gaza.

Matters reached a head after these online appeals resulted in members of the group organising an event on the borders of the Gaza Strip, calling for people to cross the border and settle there. On June 29, 2024, the group’s administrator uploaded a short clip of the event, accompanied by a post in which he thanked the security forces and local people for welcoming those who had taken part in the event, and urged the remaining members of the group to join in and lend support.

The administrator concluded his post by saying: "It is decided that this location (where the event was held) will be a permanent site, until they allow us to enter the Gaza Strip and settle there."

Event calling for renewed settlement in Gaza

This investigation was completed with support from ARIJ