Generative AI could rewrite history, including that of the Holocaust

A UNESCO report published today warns that unless decisive action is taken to integrate ethical principles, AI could distort the historical record of the Holocaust and fuel antisemitism. The report cautions that not only can Generative AI enable malicious actors to seed disinformation and hate-fueled narratives, but it can also inadvertently invent false or misleading content about the Holocaust.

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

Published in partnership with the World Jewish Congress, the UNESCO report argues that as learners increasingly use Generative AI to complete assignments and find information online, they risk exposure to distorted information about the Holocaust, which has found new ways to spread through AI-generated content. With four in five (80 percent) of young people between the ages of 10 and 24 now using AI several times a day for education, entertainment and other purposes, action to ethically guide these new technologies must be taken quickly.

Generative AI must be trained using vast amounts of data. This data is often mined from the Internet and may include misleading or harmful content. AI systems therefore inherit human biases, potentially misrepresenting information about specific events, reinforcing prejudices. This is particularly true in the context of Holocaust, because of the prevalence disinformation about this event. The report notes that due to lack of supervision, guidance and moderation by AI developers, generative AI tools may also be trained on data from Holocaust denial websites.

AI has also been documented to enable bad actors to distort Holocaust-related content, creating fabricated testimonies, and even altering historical records. Deepfake images and audio content created using Generative AI are particularly convincing for young people, who may encounter them on social media platforms. The Historical Figures App allowed users to chat with prominent Nazis such as Adolf Hitler and Joseph Goebbels, and falsely claimed that individuals such as Goebbels were not intentionally involved in the Holocaust and had tried to prevent violence against Jews.

“Hallucinating” history

Generative AI models are prone to inventing or “hallucinating” events, personalities and even historical phenomena when they do not have access to sufficient data. The report underscores that ChatGPT and Google’s Bard, have both produced content detailing Holocaust-related events which never took place. ChatGPT entirely fabricated the concept of ‘Holocaust by drowning’ campaigns in which the Nazis drowned Jews in rivers and lakes, and Bard generated fake quotes from witnesses to support distorted narratives of Holocaust massacres.

In addition to the risk of manipulations and “hallucinations” undermining established facts and trust in experts, the report also underscores the dangers of AI’s tendency to oversimplify complex history, privileging a narrow range of sources and a small selection of events. On search engines, which are powered by AI, 60-80 % of the top image results are of a single Holocaust site, Auschwitz-Birkenau.

UNESCO is calling on governments to accelerate the implementation of the Organization’s Recommendation on the Ethics of AI, the first and only global standard in this area, unanimously adopted by its Member States in 2021. This Recommendation is already being integrated into the legislation of over fifty countries.

UNESCO urges tech companies too to implement its standards, and to shoulder their responsibilities and ensure that principles such as fairness, transparency, human rights and due diligence are built into applications at the design stage. In February 2024, eight tech companies were the first to sign a commitment to the Director General of UNESCO to apply an ethical approach to the development of new AI tools.