City workers clean the “Wall of Justice” (Mur des Justes), marked with red handwritten graffiti, outside the Holocaust Memorial in Paris on May 14, 2024. The memorial was vandalized overnight, an act condemned by the president of the Council of Representatives of Jewish Institutions in France (CRIF) as anti-Semitic. (Photo by Antonin UTZ/AFP)
The United Nations warned on Tuesday that AI technology is helping to create false narratives about World War II atrocities, including Holocaust denial, and risks an “explosion of anti-Semitism”.
UNESCO, the United Nations educational and cultural agency, called on governments and technology companies to put in place ethical safeguards around AI technology and to inform schools about the risks of AI-generated content.
The UNESCO report highlighted cases where hackers had modified chatbots to spread Nazi ideology, as well as cases where bots had created their own narratives about the Holocaust.
“If we allow the horrific facts of the Holocaust to be diluted, distorted or falsified through irresponsible use of AI, we risk an explosion of anti-Semitism,” said Audrey Azoulay, Director-General of UNESCO.
She said it could also lead to a “gradual erosion of our understanding of the causes and consequences of these atrocities.”
Content-generating AI tools such as ChatGPT and DALLE-2 have exploded in 2022.
Businesses and institutions are rushing to deploy AI for everything from translating text to providing customer service.
But policymakers, activists and law enforcement have warned from the start that these tools could be used to extort money, tarnish reputations, influence elections and distort history.
Several tools have already publicly failed in history.
Google’s Gemini model sparked global controversy in February when it produced ethnically diverse images of Nazi soldiers.
UNESCO also noted that ChatGPT completely invented the concept of a “Holocaust by drowning” and that Google’s Bard chatbot fabricated eyewitness accounts to support falsehoods about Nazi genocide.
‘Red Line’
UNESCO issued a policy paper in collaboration with the World Jewish Congress to mark the International Day against Hate Speech, calling for “urgent action” from governments, technology companies and educators.
Karel Hracapane, a UNESCO expert on Holocaust education, told AFP that the distortion of Holocaust history showed how AI could upset our relationship with the truth and “lead to a serious erosion of democratic culture”.
Fracapane said the current growing popularity of far-right politicians in Western countries is part of the same process as the spread of hate speech online.
He said that while education about the Holocaust has improved significantly in recent years, the “red line” of Holocaust denial and spreading Nazi ideology has been crossed, both online and in the real world.
“The contents of this report are a reflection of what is going on in society and have very real political implications,” he said.
Among the examples highlighted in the UNESCO report were an AI-generated version of Joseph Goebbels claiming to have tried to save Jewish lives, and a deepfake of “Harry Potter” actress Emma Watson reciting Adolf Hitler’s “Mein Kampf.”
However, the UNESCO report also points out positive uses of AI.
Algorithms help sort and organize the testimonials, finding emerging patterns and fresh insights.
The UNESCO report also said that AI-based educational tools could provide immersive experiences for young people.
But Fracapane said the current state of technology is leaning more toward “AI is a threat” than “AI is an opportunity” in the debate.
©Agence France-Presse