Germany Is Using AI to Erase Pro-Palestinian Speech

1 day ago 4

In mid-February, the United Nations Special Rapporteur Francesca Albanese was scheduled to give a talk together with Forensic Architecture founder Eyal Weizman at Berlin’s Free University. Yet, the school soon faced political pressure from Israeli ambassador Ron Prosor and Berlin’s conservative mayor Kai Wegner, who demanded that its leadership “cancel the event immediately and send a clear message against antisemitism.” The university then did call off the talk, vaguely citing “security concerns.” The left-wing newspaper Junge Welt eventually offered its editorial spaces as an alternative venue.

The event went ahead under huge police intimidation: two hundred armed and riot-clad officers surrounded the building, with an additional police presence in the newspaper’s offices to ensure that no thought-crime be committed. In the days before and after, German legacy media tried not to emphasize that government officials’ intervention in the university’s affairs might threaten academic freedom. The focus was all on not platforming antisemitism: implicitly accusing Albanese and Weizman of this very offense.

These open attacks on the Palestinian diaspora, their supporters, UN representatives, and NGOs are not unique to Germany. Both US and British media often cite the claims of civic bodies like the Anti-Defamation League, the Board of Deputies, the Community Security Trust, and other anti-antisemitism organizations. The German media landscape, especially public broadcasters, more often refer to antisemitism “experts” in the form of academic scholars or government-appointed antisemitism commissioners. They are habitually presented as independent witnesses standing outside political discourse or even academic debate. Instead, their assessments — or rather, their accusations of antisemitism — are presented as objective scientific fact, not to be challenged.

A prime example of this is an interview by the Berlin newspaper Tagesspiegel with linguist and antisemitism scholar Matthias J. Becker in the wake of the cancellation of Albanese’s talk at the Free University. In it, he accuses Albanese of comparing Israel’s policies in Palestine with that of the Nazi regime and alleges that she failed to condemn the October 7 attacks.

While the latter accusation is simply false (Albanese did denounce attacks on civilians) the former claim is stated without evidence. The shrillest accusation, however, is that Albanese is projecting the “blood libel” onto Israel. This refers to a medieval antisemitic canard that charged Jews with the ritual murder of Christian children. While this accusation is used against anyone pointing out the scores of children killed by the Israeli military, it is primarily deployed by Israel’s lobbyists and spokespeople, and rarely by serious academics.

Becker was consulted by the Tagesspiegel because of his affiliation with the Decoding Antisemitism project at the Center for Research on Antisemitism at the Technical University Berlin, which he led from 2019 until 2025.  With the help of a large language computing model, the project aims to create “an [AI] algorithm that will automatically recognize antisemitic statements in web comments . . . so that antisemitic posts can be removed more efficiently and accurately” by online platforms. In a talk to the Institute of the Study of Global Antisemitism and Policy, Becker lays out the political focus of the project:

What we are interested in is not so much the antisemitism of the alt-right or white supremacy platforms but mainstream society, because . . . antisemitism on campuses, antisemitism coming from the left, among artists, it is actually the mainstream, the political moderate discourse that is a challenge in itself. . . . Because as soon as antisemitism is communicated . . . in implicit ways there is very often a lack of sanctioning in contrast for [sic] examples of antisemitic tropes uttered by a neo-Nazi.

While officially claiming to focus on “the mainstream,” the project primarily centers on Israel-related, or “new,” antisemitism. Of the roughly 103,000 individual online comments that have been collected to train the algorithm and made available as metadata on the project’s website, two-thirds relate to Palestine and Israel, while one-third concern other antisemitic incidents covered in the media. A subset comprising 21,000 comments collected immediately after the October 7 attacks found roughly 2,400 antisemitic incidents, or 11.7 percent. Almost half of these are categorized as “Attacks on Israel’s Legitimacy.”

The dataset is divided into labels of differing forms of supposed antisemitisms such as “analogies with Nazism,” fascism, apartheid, or colonialism; calling Israel a racist or terrorist state; accusing it of genocide; referencing the Boycott, Divestment and Sanctions movement (BDS); giving Israel the sole blame for the plight of the Palestinians; applying double standards; and denying Israel’s right to exist.

Albeit using publicly available data, the Decoding Antisemitism project has not itself released the full dataset. Hence, it is impossible to understand in detail which comments were categorized and archived as antisemitic and why. Further, it forecloses any holistic analysis or checks and balances of the data from outsiders (usually done via peer review in academia). However, last November the project published “A Guide to Identifying Antisemitism Online,” a five-hundred-page glossary that lays out each label in some detail. It gives examples of explicit and implicit antisemitic comments, followed by non-antisemitic comments as neutral points of reference.

Under the label “Nazi Analogy/Fascism Analogies,” a clear example of an antisemitic statement is given: “Jews are doing what Hitler did to them.” An example of implicit antisemitism reads “you stand against anti semitism and the holocaust but not against killings of innocent Palestinians. You are woman of double standards and a disgrace!” [all sic].

According to the glossary, this is antisemitic because “an implicit equation [is] being established” between Israel and Nazi Germany. This argument becomes puzzling considering the example provided of a non-antisemitic comment: “Studying the Holocaust should be a warning against all forms of oppression and injustice, whether in the Middle East or in other conflicts.” It is not clear where the substantial difference between the latter two examples lies, aside from politeness, civility, and a suspected orientation toward Zionism. Both demand consistency in the lessons learned from the past.

The most egregious instance of this approach can be found in the chapter called “Blood Libel/Child Murder.” The argument goes that accusing Israel of killing Palestinian children is akin to the antisemitic fantasy of accusing Jews of the ritualistic murder of Christian children. It is not explained how this vile trope is somehow equivalent to accusations against the Israeli military. As proof of this crude thesis, the chapter provides an example of explicit antisemitism, which reads “What you mean is, Israel bomb [sic] children. Let’s not mince words here,” whereas an example of implicit antisemitism reads “How many rockets has Israel fired on innocent children???” Perhaps, the commenters could believe in the blood libel trope. Yet clearly, neither of these comments is any kind of proof of that. How these comments are substantially different, and how a medieval antisemitic fantasy relates to a present-day genocide in real life, goes unexplained.

The possibility that someone who speaks ill of Israel and its actions could be an antisemite is turned into a must be simply because of how emotional the admonishment of Israel’s actions is perceived as being — or how political the outrage is. Unsurprisingly then, the non-antisemitic example provided of how to comment about the killing of Palestinian children is set in the passive voice: “Nine children died in Gaza last month as a result of air strikes.” According to the author, this is not antisemitic because “the statement does not suggest any deliberate action, focusing its attention on the tragic deaths” and “Choosing the verb ‘died’ instead of ‘were killed/murdered’ . . . additionally reduces the level of emotional intensity. . . .”

On the topic of Palestine and Israel, the glossary seems to operate within a logic that sees emotional responses to a live-streamed genocide not as a human reaction but as an indicator of antisemitic beliefs. The emotional, irrational, and racialized Palestinian Other is already a common theme in the criminalization of the Palestinian diaspora and Palestinian solidarity in Germany and beyond. This logic regarding the possible motivation behind innocuous or harsh comments reflects a worldview that suspects antisemitism everywhere, especially where it does not exist.

Becker makes this explicit in a seminar talk introducing the project, where he uses the analogy of the iceberg where the visible tip is the antisemitism that we can understand and identify today, but the vast majority of the antisemitism is apparently under the surface and can only be deciphered with the right methodological approach that has yet to be developed. The belief that most antisemitism is hiding under the surface, undetected and unseen, hints at a paranoid predisposition — a phenomenon all too common in German political culture but, more worryingly, also in its academic culture.

Arguably, this seemingly paranoid turn originated in German postwar antisemitism scholarship, or more precisely in the distortion and uncritical adulation of some of such scholarship. The most influential theory often used today to justify the criminalization of Palestine solidarity and to deem opposition to Israel as antisemitic is called Umwegkommunikation: detour communication. In a talk to the UCLA’s Shoah Foundation, Becker emphasizes that detour communication is part of the conceptual framework of the Decoding Antisemitism project.

Detour communication originated in 1986 when sociologists Werner Bergmann and Rainer Erb wondered — quite rightly — where all the antisemitism that had been so prevalent and institutionalized during the Nazi era had gone, once the Third Reich was dismantled. Antisemitism had been banished from the public sphere virtually overnight and what was once part of everyday political discourse had not only become a taboo but also a matter of criminal law.

Despite this, most scholarly work in Germany and the West focused on the historical emergence of an annihilationist antisemitism in early-twentieth-century Europe, which found its catastrophic conclusion in the Holocaust. These discussions were led most prominently by Theodor W. Adorno and Max Horkheimer and until Bergmann and Erb little scholarship was interested in postwar antisemitism. Surveys of postwar West-German society showed that antisemitic attitudes had not changed much but were simply not uttered publicly anymore. Hence, Bergman and Erb observed that antisemitism had morphed from an institutionalized phenomenon to a latent, hidden one that only reemerged under specific circumstances and conditions.

They theorized that antisemitism is not only a psycho-social and cultural phenomena but also a communicative one. Because antisemitism was a taboo in public discourse, other ways had to be found to utter antisemitic views without suffering the social consequences of this transgression. One way would be to ensure that the social environment of the speaker is aware and accepting of one’s antisemitic beliefs so that there is no taboo that needs breaking — primarily meaning, in private settings.

In public settings, the speaker is forced to use coded language, which ensures that the antisemitic intent of their speech is not identified as such because there is no certainty that the transgression of the taboo will be tolerated rather than sanctioned immediately. In this sense, the concept of coded language is not dissimilar to that of racist dog whistles. Yet, in the context of antisemitism, Bergmann and Erb suggest that one such code could be to speak about Israel when one cannot openly mention Jews. In this context, speaking negatively about Israel then is a strategic detour communication in lieu of openly targeting Jews.

What was set out by Bergmann and Erb as theoretical reflections on postwar antisemitism that needed to be rigorously tested was gradually taken on uncritically by other German scholars of antisemitism and over the last three decades or so morphed from a theoretical concept to a shibboleth and unproven scientific fact of Germany’s many antisemitism scholars, appointed antisemitism czars, Israel lobbyists, and other so-called experts, despite unsatisfactory empirical proof. It has become the go-to argument for why anti-Zionism is in fact antisemitism and radical critiques of Israel are a code hiding antisemitic beliefs.

Since 2019, Umwegkommunikation is listed in the entry on Israel-Related Antisemitism on the website of the Federal Agency for Civic Education and its regional counterparts, a state-funded civic education institution that provides education material for civic organizations and schools. The entry, authored by Professor Lars Rensmann, explains Umwegkommunikation in a distorted version of the original work, and goes further adding, “[V]erbal antisemitic detour communication is also suitable for legitimizing and unleashing direct violence against Jews” — something that Bergmann and Erb neither claimed nor investigated.

While scholarship on postwar antisemitism ought to be taken seriously, Bergmann and Erb’s discussions on Umwegkommunikation are overshadowed by their weaponization in attempts to claim that their work proved that anyone speaking ill of Israel or opposing its policies does so out of antisemitic motives. Additionally, the nearly forty-year-old concept has largely become obsolete in the fight against antisemitism at a time when public support for Israel has become a hallmark of the far right and of prominent corporate, political, and evangelical leaders who spread antisemitic conspiracy theories, materially support and embolden the far right while posing for photo-ops in Auschwitz and Israel, all in the name of “Never Again.”

Yet, this obvious fact is largely ignored so as not to hurt the concept’s political usefulness among German state-appointed experts, scholars, and law enforcement, who claim it as scientific fact and use it to demand ever more draconian sanctions and regulations against critics of Israel. Since the far right has recognized Israel as a blueprint for its own ethno-supremacist fantasies, using the concept of Umwegkommunikation to make sense of antisemitism has turned it from a tool to conceptualize and investigate antisemitism into one to fight the Left while ignoring or outright shielding the far right.

The Decoding Antisemitism project is the prime example of where this logic leads. It is, to date, the most authoritarian attempt to use antisemitism scholarship to erase from the public domain not only uncomfortable opposition to Israel but millions of Palestinian voices. For now, it is not entirely clear where the Decoding Antisemitism project will go next or who exactly will make use of the data it collected and the large language model it trained. My own request to access the raw dataset was denied because “the financial value of the annotated dataset has become a factor” in not publishing it — despite publication being good scientific practice.

In an interview with Israeli news outlet Mako, Becker suggests that social media providers are opening their doors and hearing concerns like his. This strongly suggests hopes to commercialize and implement its findings with online platforms. Five years after its inception, it appears that its conceptual framework and glossary have been overtaken by reality. We are today seeing, in real time, what Masha Gessen called the liquidation of a ghetto, and the deliberate killing of Palestinian children by the Israeli military, turning Gaza into a “‘Graveyard’ for children.” While this reality may shatter the scientific credibility of Decoding Antisemitism and its conceptual framework, the project may nonetheless be a formidable weapon for those who want to erase Palestinians’ and their supporters’ voices online and take legal proceedings against them offline.

Ultimately, the Decoding Antisemitism project is no aberration. It is the latest authoritarian adventure emblematic of an academic field that is often guided by a provincial hysteria and paranoia closely aligned with German foreign policy objectives and more concerned with protecting Israel’s reputation than fighting actual antisemitism. More importantly, however, this discipline has over the last four decades facilitated like no other the dehumanization of Palestinians. It has declared their lives, their suffering, and their subjugation, necessary to what is called the fight against antisemitism. In so doing, this scholarship has become part and parcel of a political culture complicit in genocide.

Read Entire Article