Data-driven artificial intelligence (AI) technologies are progressively transforming the humanitarian field, but these technologies bring about significant risks for the protection of vulnerable individuals and populations in situations of conflict and crisis. This article investigates the opportunities and risks of using AI in humanitarian action. It examines whether and under what circumstances AI can be safely deployed to support the work of humanitarian actors in the field. The article argues that AI has the potential to support humanitarian actors as they implement a paradigm shift from reactive to anticipatory approaches to humanitarian action. However, it recommends that the existing risks, including those relating to algorithmic bias and data privacy concerns, must be addressed as a priority if AI is to be put at the service of humanitarian action and not to be deployed at the expense of humanitarianism. In doing so, the article contributes to the current debates on whether it is possible to harness the potential of AI for responsible use in humanitarian action.