A keynote given at Focus on Open Science, Budapest, November 6th, 2025 Abstract: "The ongoing hype surrounding the newest generation of AI methods -- notably generative AI -- creates enormous pressure both for individuals and institutions not to get left behind in terms of technology and competence. This makes it easier for commercial tech companies to push AI-based solutions into their services, sometimes even before they have reached a certain maturity and not always in the best interest of their users. Both the training processes and productive operations of many of those solutions are using up a staggering amount of resources and creating or aggravating all kinds of risks for various minorities, the climate, and the information landscape. The full extent of the impact is a moving target and therefore hard to gauge but even the most concrete and imminent implications have barely entered the awareness of the wider public to date. In this talk we will survey a range of aspects how the use of large AI (LAI), especially proprietary LAI, could have and indeed already has a critical influence on society and its environment. We will then ask ourselves if there are alternatives to this massive commercial push of LAI applications that follow the principles of sustainability, transparency, and openness for the benefit of everyone. Libraries as publicly funded, non-profit information infrastructure institutions are uniquely qualified to promote best practices towards a more "civilized" use of AI and also to act as reliable data providers or brokers for research and productive applications. However, in order to fulfil that role, libraries have to manage the balancing act of boldly and unapologetically representing those values on the one hand and navigating the necessities of a world shaped by big profit-oriented players on the other. This is a long-term mission which will require initiative, tenacity, and strategic networking on a global scale."