UN agency warns AI could spawn Holocaust denial surge

A United Nations agency warned that generative artificial intelligence has the potential to distort history about the Holocaust and fuel antisemitism, in a report released Tuesday.

The report — published by the United Nations Educational, Scientific and Cultural Organization (UNESCO), in conjunction with the World Jewish Congress — outlined ways AI is particularly vulnerable to spreading antisemitic content and called on AI companies and governments to implement the standards set in UNESCO’s ethical principles.

“If we allow the horrific facts of the Holocaust to be diluted, distorted or falsified through the irresponsible use of AI, we risk the explosive spread of antisemitism and the gradual diminution of our understanding about the causes and consequences of these atrocities,” UNESCO Director-General Audrey Azoulay said in a press release.

“Implementing UNESCO’s Recommendation on the Ethics of AI is urgent so that younger generations grow up with facts, not fabrications,” Azoulay continued.

Generative AI systems, which are trained on large amounts of data, are especially prone to fueling pre-existing antisemitic biases either unintentionally or because malicious actors exploit the system, according to the report. Without proper vigilance, AI systems can easily be manipulated to include data from fringe parts of the internet that promote conspiracy theories and misleading claims.

Generative AI systems can also be used to create deepfakes, seemingly realistic images about the Holocaust that could be manipulated to look real to suggest the Holocaust did not happen.

The Holocaust killed 6 million Jews, or one-third of the global Jewish population, in Europe in the early 1940s, following the Nazi Party’s rise to power in the previous decade. The report warned that AI models may focus too much on some well-known images and moments from the Holocaust and miss key information that may be more complex in nature. The report warned of the potential for “hallucinations,” or the filling in of information that the models might not have easily accessible.

The report noted that 80 percent of people between the ages of 10 and 24 use AI several times a day for education and other purposes, which could fuel the biases.