In machine learning, ‘model collapse’ is a key term that refers to problems that can arise during the training of large language models, including a loss of diversity and originality in generated outputs, excessive repetition, and amplified bias. But ‘model collapse’ is also an apt metaphor for the state of current affairs. From politics to ecology, many established systems are in disarray, and as the term ‘post-truth’ suggests, intellectual governance is no exception. Digital change is a key driver of this upheaval. With artificial intelligence rolling out across society, its impact upon these crises grows ever more significant. In 2025 Vienna Digital Cultures explores ‘model collapse’ in all its senses, inviting artists and thinkers to reflect on the disorientation and instability that characterize this moment in history.