The Facebook arrests: How Israel detains Palestinians before even they know they’re terrorists

Israel has arrested hundreds of Palestinians since the lone-wolf intifada began in September 2015, partly by analyzing social media posts. Authorities say these arrests are legitimate, but others see it as a grave violation of human rights

Su’ad Zariqat’s husband was run over and killed in an accident in Israel in 2010. Ever since, the 29-year-old Palestinian says she has been regularly posting pictures of him on her Facebook page. But in the early hours of December 2, 2015, Israeli forces arrived at her home near Hebron and arrested her. This occurred at the height of the wave of violence that began two months earlier in the West Bank and Israel. It was the second time she had been arrested, after serving a prison term in 2008 over charges of contacts with organizations hostile to Israel. This time, she says, she was handed a screenshot of a Facebook post featuring a picture of her husband, with the text, “May God unite us in heaven.”

Zariqat says the word shahda (cognate to shahid, Arabic for “martyr”) in her Facebook post seemed to worry her investigator. “I told him it’s a word we use regularly. The fact that I wrote it on Facebook doesn’t mean I’ll do anything … even when someone dies in a car accident we call him shahid.”

Following the interrogation Zariqat was issued with a four-month administrative detention order – detention without trial that is used as preemptive detention by Israel, mostly against Palestinians in the West Bank. Her detention was subsequently prolonged for an additional four months. Most of the material in administrative detentions is confidential, and neither the file nor the indictment is shown to the defendant.

When asked how the detention has affected her life, Zariqat paints a bleak picture. Before she was arrested, she says, her life revolved around her desire to complete her university studies, to work and to help her family financially. Since her release, though, she has not been able to return to her studies and rarely leaves the home. Detention changed her, she says, and was traumatic.

Zariqat’s case is not exceptional. In the first months of the wave of violence that began in September 2015, many Palestinians were arrested and interrogated in connection with their activity on social networks. Since October 2015, Israel has arrested more than 200 Palestinians on charges of incitement on social media. Their lawyers describe circumstances and sequences of events similar to those that characterized the experience of the young widow. Her profile – a relative of a shahid who lives in the Hebron area – is a perfect fit for the profile Israeli security forces constructed in order to assess the threat level of individual Palestinians who might carry out stabbings and car-ramming attacks.

According to Foreign Ministry data, 43 Israelis have been killed and 682 injured in terror attacks in Israel and the West Bank since September 2015.

The UN Office for the Coordination of Humanitarian Affairs says that in that same period, 237 Palestinians were killed, including 167 who were killed while attempting to carry out attacks on Israelis. Others were killed in Israeli army raids and at demonstrations, including demonstrations held along the Gaza Strip border. The estimated number of Palestinians wounded during that period is 15,000, according to OCHA figures.

Lt. Col. (ret.) Maurice Hirsch was the chief military prosecutor for Judea and Samaria (the West Bank) until last year. He says that since September 2015, dozens of social media posts that seemingly indicated the writer’s intention to carry out a terror attack culminated in administrative detentions. In other cases, people who were identified as liable to carry out a terror attack were convicted of incitement on social media.

Hirsch explains that the military prosecution is obligated to conduct an exhaustive criminal procedure before it has recourse to the option of administrative detention. Therefore, if there is sufficient evidence for a charge of incitement, use will be made of it – even if there is a suspicion that the suspect is a potential terrorist, Hirsch adds. He notes that in most of the cases in which criminal charges were filed regarding individuals suspected of being potential attackers, this provision of the law was used.

Sami Janazreh, 43, was placed under administrative detention in November 2015. He was arrested at his home in the Al-Fawwar refugee camp, near Hebron. He says that initially he didn’t understand why he was being arrested. As with every administrative detention, most of the material in his case file was confidential and neither the file nor the indictment were shown to him. Shortly after he was detained, he decided to go on hunger strike, demanding to know what the charges against him were.

After 71 days on hunger strike, Janazreh’s fight succeeded – at least partially. Talking to Haaretz from his Al-Fawwar home, he says he was told that instead of administrative detention, he would be standing trial for incitement on social media.

“My indictment was made up of screenshots of my Facebook posts,” he says. “And then I realized that the war against Facebook arrests is exactly like the war on administrative detention – any Palestinian today could be guilty. Every Palestinian whom the Shin Bet security service finds has shared a picture of a shahid or a prisoner, or has written a Facebook post about himself as a Palestinian – they could say this is incitement.”

Dr. Itamar Mann, a lecturer on international law and political theory at the University of Haifa, says it is not by chance that these practices – incitement trials and predictions on the basis of social media activity – exist in Israel’s military courts in the West Bank (the system through which Israel tries Palestinians in the West Bank).

“From the perspective of the legal system,” he says, “there is a difference between the way Israel allocates rights within Israel and the way it allocates rights in the Palestinian Territories. And specifically, there is a difference in the limits on the right to freedom of speech. Human rights laws do not apply in the Palestinian Territories and therefore freedom of speech is not given the same protection in the military courts.”

According to Palestinian organization the Addameer Prisoner Support and Human Rights Association (a Palestinian advocacy nonprofit organization), more than 800,000 Palestinians have been arrested under Israeli martial law since 1967 – that is, 20 percent of the total Palestinian population and 40 percent of adult males.

Generalized profile of potential attackers

Since the wave of violence (also known as the lone-wolf intifada) began in September 2015, the Israeli authorities developed an early alert system that assesses the likelihood of the involvement of individual Palestinians in terror attacks. In July 2016, the IDF Spokesman’s Office conducted several press briefings, in which the system was described by an Israeli intelligence officer familiar with the matter. In recent requests, the Israeli army denied the authors of this article access to relevant interviews, citing policy changes.

In April, Haaretz’s Amos Harel reported that in little over a year, this method identified some 2,200 Palestinians in various stages of the decision to carry out terror attacks or in the planning of such an attack. Some 400 were subsequently arrested by the Israeli army and the Shin Bet. The names of about 400 additional individuals were handed to the Palestinian Authority and they were arrested by security organizations in the West Bank and given warnings.

“Unlike terrorists who belong to Hamas or the Islamic Jihad, if you get to their house a week before the attack the kid doesn’t know that he is a terrorist yet,” noted an officer in a briefing held at an IDF base in July 2016. Haaretz is in possession of a recording of this briefing.

Soon after the beginning of the wave of attacks, Israeli authorities assigned dozens of intelligence officers to assemble generalized profiles of “potential attackers,” according to the officer. Initially, three or four generalized profiles of the early attackers were constructed, combining data points including age, location of residence and an assessment of the psychological build-up and intentions of the attackers, based on the information available to the authorities. The latter included social media posts, and intelligence from further sources. Some of the attackers were also interrogated in the process, the officer said.

Based on these assembled profiles, possible would-be assailants were assessed to be mostly under 25 years old, around 40 percent of them going through personal hardships, and presumably wishing to become martyrs as an honorable way to commit suicide. Other data points include security records, if there are such, and a mapping of terror-related activities by the person’s close relations, including family, according to the briefing. Personal problems, including family tensions and forced marriages, were described by the officer at the July briefing as being strong motives, especially with female attackers. Israeli authorities identified several West Bank villages and neighborhoods in East Jerusalem as the source of around half of the attacks.

Nearly daily “risk management sit downs” are held by Israeli intelligence officers to decide on the preferred course of action with regard to the individuals flagged by the system, according to the officer. “You still need an analyst … you can’t say, ‘OK, I’ll just arrest everyone who is 16 years old and has [suicidal tendencies] and he is from the village that you had attacks [coming from]. … You can’t arrest everyone who just had some sort of question in his head that he might want to stab a soldier,” he said.

As noted by Israeli journalist Ehud Yaari in The American Interest in January, special efforts were devoted to hacking into messaging apps in order to broaden data collection. “People today change their ways of communication every week, so you have to be very, very broad in the ways you collect data and on the specific information that you’re searching for,” said the officer in the July briefing.

In addition to monitoring social media posts, Israeli authorities also acted to remove online content inciting to violence, as Yaari noted in his article. The ensemble of these activities constituted a major shift in resource allocation at the Shin Bet, Yaari added, resulting with an assignment of up to a third of the agency’s personnel to the technology division. This reform was described in detail by Amos Harel in an April report in Haaretz (Hebrew edition).

During the first months of the violence in 2015, the number of Palestinians unaffiliated with terror organizations who were held in Israeli jails jumped 66 percent – from 648 arrests to 1038, according to Israel Prison Service data. The figures provide an indication of the changes that have taken place in the population of prisoners held in Israeli jails due to security concerns since September 2015.

It is reasonable to assume it would also be possible, with relevant adjustments, to make use of the same system in order to identify Jewish would-be attackers. But according to the July 2016 press briefing, it was not used in such a way either in Israel or the West Bank.

The Facebook prisoners

Both lawyers and former detainees interviewed for this article describe the “Facebook prisoners” as ones matching the criteria described by the officer at the July briefing – most of them young, either from the Hebron area, from refugee camps or from Jerusalem, and without previous detention history. “During my detention I met other Facebook prisoners, but most of them were on administrative detention. … They target people from Hebron the most,” Yousef al-Jaabri, 20, tells Haaretz. He was imprisoned for six months on charges of incitement on Facebook.

Between October 2015 and the end of 2016, some 160 to 170 incitement cases concerning social media were brought before military courts in the West Bank, according to court records and military prosecution personnel. Most of the Palestinians who were arrested on suspicion of incitement were held in detention for between six to 18 months, and in cases where an indictment was filed, they were released in plea bargains.

A further 60 cases of incitement, most of them social media-related, were brought before civil courts in Israel during that period – compared to only 30 cases between 2011 and 2014, according to Israeli Justice Ministry figures.

“We reached a stage in which ‘What’s the name of your Facebook profile?’ became the first thing which is asked in investigations of [Palestinians] arrested in the West Bank,” says Fadi Qawasmi, a lawyer who represents many Palestinians in the West Bank’s military courts. “We then see this [Facebook incitement] as a count added to the charge of stone-throwing, for example, or as a single charge.”

One of the basic building blocks that enabled the arrest and conviction of many Palestinians during this period can be found in a case held in the Ofer military court in February 2016, to which was appended an expert report by a Shin Bet official. The report asserts that 70 percent of assailants who had social media accounts had “extreme and irregular” expressions on Facebook.

Former military prosecutor Hirsch says he explained to the prosecutors and investigative authorities that he likens social media “to a person standing on a box in Hyde Park. Does the fact that we are talking about social media make such a big difference?” According to Hirsch, from the prosecution’s perspective, a person who “likes” something on Facebook is like someone in the audience who stands there and nods his head. However, a person who shares the content “brings himself closer to the level of the inciter himself, because he broadcasts the entirety of the inciting content.”

Posts presented to court as evidence of incitement on Facebook often cite Koranic verses or present photos of martyrs. Some of the posts can be seen as statements testifying to suicidal tendencies, or the writer’s intention to sacrifice his life. “My will, if I never come back, is that I’ll meet you in heaven,” wrote a 17-year-old from Hebron in one of the posts included in an indictment for incitement.

Other posts presented in court were clearly political in nature – some referred to the Al-Aqsa mosque, while others dealt with resistance to the Israeli occupation. Some of the posts included a direct call for violence against Israelis. While some of the Palestinians who were convicted on charges of incitement on Facebook have a considerable audience – sometimes thousands of friends – in other cases the accused had less than 50 friends and their posts had as few as two “Likes.”

Dr. Yonatan Mendel, a scholar of the Arabic language and author of the article “The Politics of Non-Translation: On Israeli Translations of Intifada, Shahid, Hudna and Islamic Movements” (published in 2010 in the Cambridge Literary Review), says the Israeli perception of incitement is nourished in part by the one-dimensional understanding of political and religious terms the Palestinians use – referring to the fact that the Facebook posts and overall evidence is translated from Arabic to Hebrew by the military court.

“Very often there is an incriminating translation that isn’t faithful to the political and linguistic context in which it was written,” he says. “For example, in the Israeli mind, a call for intifada is a call for violence. However, if you say in Arabic ‘I am calling for intifada,’ the meaning is to oppose the violence directed at us and therefore has a lot more depth. Even going out to demonstrate is intifada, and nonviolent resistance is also intifada. Similarly, when Palestinians use the word ‘shahid,’ they first and foremost are referring to the concept of ‘victim.’ Even a person who dies of a heart attack at a checkpoint is a shahid, and children who died in bombings are shahids.”

Jews incite, too

In May 2016, Arif Jaradat, a 23-year-old Palestinian with Down syndrome, was shot and killed by IDF soldiers in his village of Sa’ir, near Hebron. His family says he was shot after he started shouted and walking toward a group of soldiers that had entered the village.

Two months later, the IDF arrested his brother, Hiran, during a night raid on his home. Hiran told Haaretz that his interrogators showed him screenshots of posts he had published on Facebook that included photographs commemorating his brother.

Were you asked explicitly if you were intending to carry out a terror attack?

“When they showed me the picture of my brother Arif, they said to me, ‘Maybe you want to avenge your brother’s death and carry out a terror attack?’ I said ‘No, I love life and don’t want revenge because of my brother or anyone else.’”

And how does your life look after the arrest?

“With respect to the use of Facebook, I will go back to posting things there – but I will be more careful about what I post. You can’t really know if a certain picture is considered incitement or not.”

Hiran Jaradat’s indictment included 33 posts from his Facebook page. Most of them are about shahids and two of them support violence against Israelis. None include the pictures of his brother Arif.

According to a report published by the Berl Katznelson Foundation (an Israeli academic institute affiliated with the Israeli Labor Party), between June 2015 and May 2016, 175,000 calls for violence were posted in Hebrew online, 50 percent of them directed against Arabs. However, Arab suspects were involved in more than half of the 594 investigations pertaining to incitement conducted by the Israel Police between September 2015 and the end of 2016. According to Israeli police data, the number of indictments filed against Arabs was nearly three times the number filed against Jewish suspects.

‘Borderline cases’

One of the major criticisms leveled against the use of predictive systems for police purposes concerns the possibility of suspecting people who would not have committed a crime, regardless of police intervention. In data-processing parlance, such cases are called “false positives.” Guy Caspi is the CEO of Israel-based Fifth Dimension Holdings Ltd., a company developing predictive analytics systems used by security agencies in Israel and abroad. False positives are an integral part of the operation of computerized forecasting systems, he says, adding that there are currently only two technology companies that provide predictive systems used for intelligence at Israeli security organizations – Fifth Dimension and the Palo-Alto based industry leader Palantir Technologies Inc. Palantir has a sales and deployment staff with offices located in Tel Aviv. Palantir did not comment.

In 2015, after he retired from active service, former IDF Chief of Staff Benny Gantz was appointed chairman of Fifth Dimension. In 2016, the company appointed Ram Ben-Barak, formerly deputy head of the Mossad, as its president. According to Caspi, the company’s clients can determine by themselves the proportion of false positives trading cases identified as undesirable while marked as desirable versus cases to be marked as desirable while they are undesirable. In the case of financial transactions, for example, such a trade-off would be between fraudulent transactions deemed legitimate and legitimate transactions deemed fraudulent.

“Intelligence organizations say: ‘I don’t care about 2 or 3 percent of false positives,’” says Caspi. He adds that Fifth Dimension develops decision-support tools, and that policing actions taken with regard to the system’s output are decided upon by human personnel at the security agencies.

Referencing Steven Spielberg’s 2002 sci-fi film “Minority Report,” about a police unit that uses foreknowledge to stop future crimes, he says: “We aren’t there yet. Someone who is examined by the system isn’t automatically judged and you bring him before a court.” He adds that compared to intelligence organizations, the sensitivity to false positive cases in the financial sector is much higher.

Intelligence Affairs Minister Yisrael Katz – who also serves as a member of the security cabinet – confirmed to Haaretz at his office in Tel Aviv that there is a possibility some of the Palestinians who were arrested after having been marked out by the predictive system were not actively and fully planning to carry out an attack, and perhaps did not decide to carry out an attack at the time of their arrest. Katz says this may happen in “borderline cases.”

“As a result of the unique system that was developed and put into operation here, hundreds of cases of attacks of this sort have been prevented. In the dilemma of whether or not to act – it could be that you also include borderline cases here,” Katz adds.

In March 2016, a 17-year-old Palestinian girl from a village near Jenin was arrested as she was in a taxi on her way to Tapuah Junction, the scene of many attempted attacks. Border Police officers stopped her after receiving a specific intelligence warning about her. She was searched and found to be carrying a knife, ostensibly for the purpose of conducting an attack. According to the July 2016 briefing, an alert concerning her was sent after she was flagged by the computerized system. The briefing officer noted that she popped on the system’s radar because of indications that she was having problems with her parents and that she might have been suffering from depression.

After a peak of 80 attempted attacks in October 2015, the number of attacks has steadily declined. Since April 2016, the number of attempts has dropped to less than 20 a month, according to Foreign Ministry data. Katz sees this decline as clear evidence that the methods used by Israel to combat attacks had taken effect, including the use of the early-alert system.

“Arrests were made, terror attacks went down, that means these are the folks, no way around it. Statistically, these are the [right people]. If we missed this way or that way in a few cases? The cause is deserving – preventing terror attacks. It is not as if someone invented a way to haunt someone on Facebook.”

The cost and the benefit

The combination of a computerized predictive system and the indictment mechanism as a way to battle lone-wolf attacks is unique to the State of Israel, but the use of such technologies is growing steadily among police forces and security agencies worldwide. Technology firms selling predictive policing tools include Palantir, as noted above, and multinationals like IBM and Motorola.

Two studies conducted by the RAND Corporation (a nonprofit think tank funded by the U.S. government), found that computerized predictive systems used by the police to have no actual effect on public safety. In one of the RAND studies, the researchers concluded that an early version of the system used by the Chicago police did not reduce the number of shooting incidents in the city. In another study, the researchers found no statistical evidence of a reduction in crime in the city of Shreveport, Louisiana – which is where a computerized crime-prediction system was first used.

The researchers found that without guidance about the proper use of predictive data, police officers in Shreveport “stopped individuals who were committing ordinance violations [e.g., walking in the middle of the street]” in predicted crime hot spots.

Human rights groups and legal scholars point to risks in the use of predictive policing tools. Beyond privacy infringements, critics fear predictive policing systems can heighten unneeded police focus on specific racial and ethnic groups, saying that given the lack of transparency involved, biased results can go unnoticed.

In raids carried out by the Chicago police in 2016, most of the people arrested had appeared in lists of potential troublemakers created by the computerized prediction system. They were arrested for weapons possession and drug dealing.

“There’s a great fear around predictive policing around the world and in the United States that these tools will lead to preemptive deprivation of liberty,” says David Robinson, a principal at policy consulting firm Upturn. “The traditional view of how this is supposed to work is that deprivations of liberty are the consequence of law breaking. The idea of depriving someone of their liberty before they’ve done anything wrong is a fundamental departure from the model policing is traditionally based upon. In any society, it is worth pausing and thinking very hard about whether to cross that bridge or not.”

The IDF Spokesman responded to this article: “In the past two years, the State of Israel has been facing a wave of terror, accompanied by severe incitement to harm Israel’s citizens and soldiers. At the same time, we witness the phenomenon of terror attacks being carried out by individuals or groups following exposure to content that incites, and who are inspired by such content.

“The security forces are waging an extensive campaign against terrorism, in order to preserve the security of the state and its residents,” the spokesman continued. “In this context, various measures are being taken in order to prevent terror attacks and to combat the phenomenon of incitement. The writers’ treatment of data that were presented in the past in a press briefing at which they were not present is inaccurate and takes things out of context. We would like to stress that the security forces are acting in order to gather and examine in-depth information that comes into their hands.

“We will add that administrative detention orders are issued when there is established security information and which necessitates arrest. These orders are issued after careful examination of the relevant information and are subject to judicial review. Indictments for crimes of incitement are filed when evidence is gathered that indicates use of severe language that incites to terror and grave violence.”

The IDF spokesman did not address a request to point out any inaccuracies in the article or to places in which things said in the briefing are taken out of context.

At the request of the military censor, certain sections of this article, about the way the early alert system works, have been removed.

The research for this article was made possible by support from


Does the European Union ‘like’ what Israel is doing?

Predictive policing has been known and practiced in several European Union member states for years. Police departments across Britain have been experimenting with predictive policing tools, according to a report by European civil rights organization Statewatch. However, the deployment of national-level, person-based predictive systems still seems to be out of regulatory reach for EU members and to contradict articles pertaining to privacy and nondiscrimination in the Charter of Fundamental Rights of the European Union.

“For counterterrorism needs – to be able to say this and that person is about to do so and so – you need a learning system that has regulatory freedom and can integrate various information streams,” says Fifth Dimension’s Caspi.

This has not prevented EU officials from expressing an interest in adopting Israeli methodologies. In conferences in Israel and in meetings with Israeli officials in 2016, the EU’s counterterrorism coordinator, Gilles de Kerchove, expressed interest in adopting Israeli technologies to battle lone-wolf attacks. De Kerchove discussed at length the EU’s plan to battle online incitement and its difficulty in finding enough speakers of Middle Eastern languages to manually monitor content.

Last October, Danish newspaper Information reported that Danish Police had bought a platform for predictive policing from Palantir Technologies. Following this purchase, this February Danish Justice Minister Soren Pape tabled a draft law for public consultation with the aim of taking broad use of data trawling to prevent crime.

Criminalizing dissemination of messages via the internet is now an EU-wide practice based on a Framework Decision on Combating Racism and Xenophobia adopted in 2008. An EU directive on combating terrorism signed in March makes it compulsory for member states to criminalize the distribution of messages ”glorifying terrorist acts.” Member states will also be obliged to remove or block provocations to commit terrorist offenses from the web.

Spain is at the forefront in convictions for “glorifying terrorism.” There were 19 convictions in 2015, and another 27 last year. Earlier this year, a Spanish court jailed rapper César Strawberry (real name, César Montaña Lehman) for a series of tweets dating back to 2013. The tweets, which César described as humorous, included a comment that he wanted to send the Spanish king a “cake-bomb” for his birthday. In César’s case, he says there was no intention that actual acts of terrorism should be committed. But the Spanish penal code makes no distinction of intent. The Supreme Court ruled that intention was “irrelevant.” (Staffan Dahllöf and Jennifer Baker)