Is AI Writing History?
Conference Report from the House of the Wannsee Conference

Anyone scrolling through social media today is increasingly likely to encounter images such as these: an emaciated elderly man wearing a striped concentration camp uniform, with uniformed soldiers standing in the background. At first glance, the image appears authentic. But how can we tell whether it is genuine or an AI-generated construction assembled from millions of historical photographs and designed to provoke an emotional response? In January 2026, more than 30 memorial institutions called for decisive action against the spread of this form of "fake history" on social media platforms.

On 17 and 18 June 2026, the House of the Wannsee Conference Memorial and Educational Site hosted the conference Is AI Writing History? with support from the Alfred Landecker Foundation. The event explored how artificial intelligence is reshaping historical research, remembrance culture and our engagement with history. Over two days, panels and workshops brought together technological perspectives with historical scholarship and ethical reflection.

Opening the conference, Johannes Rudloff, Grants Manager at the Alfred Landecker Foundation, highlighted the opportunities presented by AI-generated content: "AI can help preserve the testimony of Holocaust survivors, unlock previously inaccessible sources, process vast amounts of material and overcome barriers to access. For many people – especially younger generations – a prompt is now becoming their first encounter with history, including the history of National Socialism and its mass crimes." At the same time, he stressed the risks: AI is accustoming us to a world in which authentic and fabricated content become increasingly difficult to distinguish, threatening both the integrity of historical transmission and the reliability of survivor testimony.

Social Media as the Engine of "Fake History"

The opening keynote was delivered by Professor Felix Stalder of the Zurich University of the Arts under the title Analysis and Generation: The Truths of Artificial Intelligence. Responding directly to the conference's central question, he argued that AI does not write history – but it does write along with it. Technologies shape the horizon of what can be thought, and generative AI does so in particularly powerful ways. AI calculates the probability of statements rather than their truthfulness; genuine understanding is not part of these models. AI-generated images are therefore not representations of an external reality, but synthetic constructions derived from patterns in data. Social media, he argued, is the real engine of their dissemination – with AI acting as the fuel.

Between New Opportunities and a Loss of Trust

At the beginning of the second day, Professor Tobias Ebbrecht-Hartmann of the Hebrew University of Jerusalem identified what he considers the greatest challenge in today's use of AI: the erosion of credibility. AI can help fill gaps in archives and assist, for example, in identifying names. At the same time, however, it produces errors and statistical artefacts that complicate scholarly work. He called for taking scepticism seriously without falling into a position in which no source is regarded as trustworthy anymore.

Labelling Requirements Alone Are Not Enough

The panel Preserving, Unlocking, Passing On – AI in (Digital) Archives featured Dr Anna Menny from the Institute for the History of the German Jews, Clemens Neudecker from the Berlin State Library, Verena Lucia Nägel from Freie Universität Berlin, and Dr Alina Bothe from Freie Universität Berlin and project lead of #LastSeen. Together, they discussed the role of AI in historical research.

Dr Alina Bothe argued for a stronger political and power-critical framing of large AI models. Historical sources make meaningful statements about the past possible, she noted, and digital source criticism therefore urgently needs to evolve in order to contextualise AI-generated material reliably. Dr Anna Menny emphasised the responsibility of historians themselves: those who collect and preserve the data being produced today also shape the histories that will be written tomorrow.

The panel Fake, Fact and Act – Credibility and Truth in the Age of AI brought together Dr Iris Groschek from the Hamburg Memorials Foundation, Dr Vera Schmitt from the Technical University of Berlin, and Dr Lisa Käde from the Berlin law firm JBViniol to discuss AI-generated Holocaust imagery, the detection of disinformation and legal perspectives. The panel reached a clear conclusion: labelling requirements alone are insufficient. What is needed is coordinated, cross-institutional cooperation at the European level to develop shared guidelines, legal frameworks and clear boundaries for the use of AI.

Defending Remembrance Culture Online

The workshops, panels and lectures demonstrated that evidence-based remembrance culture must be actively defended online. AI can provide valuable support by helping to process large collections of historical data, transcribe and translate historical sources, improve access to archives and identify existing gaps in archival collections. At the same time, it opens up new possibilities for artistic and educational engagement with the history of National Socialism and the Holocaust. Such applications, however, must never be mistaken for authentic historical representations. Instead, they should be understood as interpretative approaches that require critical reflection.

The conference was organised by the House of the Wannsee Conference Memorial and Educational Site, which has been dedicated for decades to historical and civic education on the history of the Wannsee Conference and the Holocaust. Cooperation partners included the DAAD Center for German Studies at the Hebrew University of Jerusalem and kulturBdigital at the Technologiestiftung Berlin.