"Never Again" in the age of AI requires a different kind of response

Holocaust Inversion is the portrayal of Israel and Jews as modern-day Nazis. Historian Robert S. Wistrich called it the most potent contemporary form of antisemitism. The IHRA Working Definition of Antisemitism (2016) reflects this directly, identifying comparisons between Israeli policy and Nazi Germany as a defining example of the phenomenon.

It does not erase the Holocaust. It repurposes it. Israel becomes the "new Nazi," Palestinians the "new Jews," and the moral authority of history's defining atrocity is turned against the Jewish state. The purpose, as former U.S. Assistant Secretary of Education Kenneth Marcus described it, is engineered to "shock, silence, threaten, insulate."

The language comes from the top. In 2022, Palestinian Authority President Mahmoud Abbas stood in Germany, the place where the Holocaust was planned, and declared that Israel had committed "fifty Holocausts" against the Palestinian people. This week, Erdogan launched an unprecedented attack against the State of Israel, calling it "the Hitler of the 21st century." They are not outliers. They are articulating the frame that has become consensus in significant parts of Western public opinion.

A Pew Research Center survey conducted in March 2026 found that 60% of American adults now hold an unfavorable view of Israel, up 18 points since 2022. Among Democrats aged 18 to 49, that figure reaches 84%. The share of Americans with a very unfavorable view has nearly tripled in four years. This is not a shift in policy preference. It is the inversion, having arrived.

When historical memory fades, inversion fills the space.

After October 7, that dynamic became structural. The Hamas massacre was the deadliest day for Jews since the Holocaust. What followed on the international stage was not a hearing. It was a foreclosure. Rape as a weapon of war, systematic torture, the deliberate targeting of families, documented, verified, undeniable. What proved harder to shift was not the evidence but the frame. When the dominant international narrative positions Israel as perpetrator rather than victim, survivor testimony rarely finds a platform. The Israeli voice struggled to be heard. That is Holocaust Inversion in its most direct political application: not a slogan, a filter on who gets heard.

What has changed is not the logic of the inversion. What has changed is its infrastructure.

A 2024 UNESCO report, published with the World Jewish Congress, found that generative AI systems may be trained on data that includes Holocaust denial content. The consequences are documented. ChatGPT fabricated a fictional Nazi campaign of drowning Jews in rivers and presented it as historical fact. Google's Bard generated fake eyewitness testimony to support distorted Holocaust narratives. A Historical Figures app allowed users to converse with Joseph Goebbels, and falsely claimed he had tried to save Jewish lives. An AI-generated voice of Emma Watson was used to read passages from Mein Kampf. These are not edge cases. They are what happens when systems are trained on unfiltered data at scale. Eighty percent of young people aged 10 to 24 use generative AI multiple times a day.

The ADL's March 2025 evaluation of the four dominant large language models, GPT, Claude, Gemini, and Llama, found measurable anti-Jewish and anti-Israel bias in all four, with every model more biased against Israel than against Jews generally. The most precise finding: nearly all models gave more biased responses when asked whether Israelis exploit Holocaust memory for political gain than when asked the same question about Jews overall.

Holocaust Inversion is not only a narrative spread by human actors. It is the default calibration of systems that hundreds of millions of people trust to be objective.

The Islamic Republic of Iran institutionalized Holocaust denial as state policy for decades, calling Israel a "cancerous tumor" and invoking the Nazi phrase "Final Solution" with deliberate precision. It financed October 7. For 47 years it systematically repressed its own citizens — executing dissidents, using rape as an instrument of state control, massacring tens of thousands of protesters in January 2026 in what Amnesty International called the deadliest period of Iranian repression in decades of its research. And when coordinated strikes against the regime began in February, the same inversion it had exported for forty years was activated in Western discourse: the operation was framed not as a response to a state that had funded mass atrocities, but as evidence of Jewish control over American foreign policy. Within 48 hours, the ADL tracked the antisemitic conspiracy — rebranded online as "Operation Epstein Fury", spreading to over 91,000 mentions from 60,000+ unique accounts on X alone. The Blue Square Alliance Command Center documented a 749% surge in posts framing the U.S. as a puppet of Israel. The perpetrator had become the victim.

In the same tome, the Iranian protesters faced the same regime that armed Hamas, that used rape as a systematic instrument of repression, that built its power through the same playbook of violence and denial. Their testimony, like October 7 survivor testimony, is dismissed as too political to engage. The mechanism is identical: when a regime's victims are positioned as its perpetrators, testimony becomes inadmissible by design. Those of us working at this intersection have witnessed the exception, genuine solidarity between Iranians and Israelis, expressed across borders by people who share no political language but share the same adversary. That bond is not diplomatic. It is evidentiary.

The answer cannot be rebuttal. By the time a correction reaches an audience, the narrative has already landed. The answer is truth — not as a declaration, but as infrastructure.

That means two things simultaneously. First: using technology to detect and map distortion at scale — identifying where inversion is being generated, how it moves, which systems amplify it, and at what velocity it reaches which audiences. The algorithmic bias documented in large language models, the Holocaust denial content embedded in training data, the coordinated surge of antisemitic narratives that follows every major event — these are not invisible. They are measurable. They can be tracked, exposed, and interrupted. Second: ensuring that verified truth — about the Holocaust, about October 7, about the Iranian regime's record — reaches the institutions and publics that shape policy and law, before the distortion calculates into consensus.

Neither is possible without coalition. Human rights organizations with international legitimacy. Technology companies with the tools and the obligation to address algorithmic bias in their own systems. Israel's startup ecosystem, which has developed capabilities in information analysis and rapid response under conditions no other environment has faced. Academic institutions that supply the evidentiary foundation. Governments that set standards. Organizations fighting antisemitism on a daily operational basis, whose real-time data is the only early warning system that exists. The distortions are winning not because they are true, but because the infrastructure of truth has not kept pace.

Holocaust Remembrance Day will always be rooted in memory. What this year makes clear is that memory is no longer sufficient. The systems shaping how the next generation understands the Holocaust, October 7, and Iran are not neutral. They have already taken sides.

Technology is already part of the problem. Used responsibly, it can also be part of the answer.