The evolution of river systems and their related fluxes is considered at various time scales: (i) over the last 18 000 years, under climatic variability control, (ii) over the last 50 to 200 years (Anthropocene) due to direct human impacts. Natural Holocene variations in time and space depend on (i) land-to-ocean connections (endorheism, glacial cover, exposure of continental shelf); (ii) types of natural fluvial filters (e.g., wetlands, lakes, floodplains, estuaries). Anthropocene changes concern (i) land–ocean connection (e.g., partial to total runoff reduction resulting from water management), (ii) modification and removal of natural filters, (iii) creation of new filters, particularly irrigated fields and reservoirs, (iv) acceleration and/or development of material sources from human activities. The total river basin area directly affected by human activities is of the same order of magnitude (>40 Mkm2) as the total area affected over the last 18 000 years. A tentative analysis of 38 major river systems totaling 55 Mkm2 is proposed for several criteria: (i) trajectories of Holocene evolution, (ii) occurrence of natural fluvial filters, (iii) present-day fluvial filters: most river basins are unique. Riverine fluxes per unit area are characterized by hot spots that exceed the world average by one order of magnitude. At the Anthropocene (i.e., since 1950), many riverine fluxes have globally increased (sodium, chloride, sulfate, nitrogen, phosphorous, heavy metals), others are stable (calcium, bicarbonate, sediments) or likely to decrease (dissolved silica). Future trajectories of river fluxes will depend on the balance between increased sources of material (e.g., soil erosion, pollution, fertilization), water abstraction for irrigation and the modification of fluvial filters, particularly the occurrence of reservoirs that already intercept half of the water and store at least 30% of river sediment fluxes. In some river systems, retention actually exceeds material production and river fluxes are actually decreasing. These trajectories are specific to each river and to each type of river material. Megacities, mining and industrial districts can be considered as hot spots of contaminants fluxes, while major reservoirs are global-scale sinks for all particulates. Global picture should therefore be determined at a fine resolution, since regional differences in Anthropocene evolution of river fluxes may reach one order of magnitude, as illustrated for total nitrogen. To cite this article: M. Meybeck, C. Vörösmarty, C. R. Geoscience 337 (2005).