The UNESCO report, published in collaboration with the World Jewish Congress, explains that as learners increasingly use generative AI to complete assignments and search for information online, they risk being exposed to distorted information about the Holocaust. This information is spread in new ways through AI-generated content. With four in five young people aged 10 to 24 (80%) now using AI several times a day for education, entertainment and other purposes, steps must be taken urgently to ethically guide these new technologies.
AI has the potential to absorb and amplify societal biases
Generative AI must be trained using vast amounts of data. This data is often mined from the internet and may contain misleading or harmful content. AI systems may therefore inherit human biases and misrepresent information about certain events, reinforcing prejudices. This is especially true as misinformation about the Holocaust is widespread. The report notes that due to a lack of oversight, guidance and moderation by AI developers, generative AI tools may be trained using data from Holocaust denial websites.
AI has also been documented to enable bad actors to distort Holocaust-related content, create false testimonials, and even falsify the historical record. Deepfake image and audio content created using generative AI is particularly persuasive for young people who may encounter it on social media platforms. A historical figures app allowed users to chat with prominent Nazis such as Adolf Hitler and Joseph Goebbels, falsely claiming that figures such as Goebbels had no intentional involvement in the Holocaust and tried to stop violence against Jewish people.
A history of “hallucinations” caused by missing data
Generative AI models have a tendency to fabricate or “hallucinate” events, people, and even historical phenomena if they don't have access to sufficient data. The report highlights that both ChatGPT and Google's Bard produced content detailing Holocaust-related events that did not actually occur. ChatGPT completely fabricated the concept of the “Holocaust by Drowning” campaign, which claimed the Nazis drowned Jews in rivers and lakes, and Byrd produced fake quotes from eyewitnesses to support his distorted story of the Holocaust genocide.
In addition to the risk of manipulation or “hallucination” undermining trust in established facts and experts, the report also highlights the danger of AI tending to oversimplify complex history, prioritising a limited number of sources and a few events: 60-80% of the top image search results on AI-powered search engines are images of a single Holocaust site, Auschwitz-Birkenau.
Solution: Take decisive action to put UNESCO's ethical principles into practice
UNESCO is calling on governments to accelerate the implementation of its Recommendation on the Ethics of AI, the first and only global standard in this field, unanimously adopted by Member States in 2021 and already incorporated into the laws of more than 50 countries.
UNESCO also urges technology companies to implement its standards, hold them accountable, and ensure principles such as fairness, transparency, human rights, and due diligence are built into their applications at the design stage. In February 2024, eight technology companies signed the first ever pledge to the UNESCO Director-General to apply an ethical approach to the development of new AI tools.
Technology companies should work closely with Jewish communities, Holocaust survivors and their descendants, educators, experts on anti-Semitism, and historians as they develop new AI tools.
UNESCO also calls on education systems to support the preservation of the truth of the Holocaust by equipping young people with digital literacy, critical thinking skills and an accurate understanding of the history of this genocide.
