The analysis of CO2 and its stable carbon isotopes from ice cores revealed large changes of atmospheric CO2 which are closely related to a reorganisation of the global ocean circulation, marine processes and minor contributions in the terrestrial carbon storage. These components dominate the large CO2 amplitudes during glacial/interglacial terminations. Yet, on longer orbital time scales,CO2 is also modulated by the alkalinity of the ocean system. The net alkalinity influx to the ocean is driven by silicate weathering, which draws down atmospheric CO2 and provides alkalinity in the form of bicarbonate ions. Conversely, alkalinity is lost during coral reef growth and when CaCO3 is buried in marine sediments. On orbital time scales, these fluxes are assumed to be almost balanced as atmospheric CO2 and its climatic effects feed back on the weathering rates providing a negative feedback loop. Besides these basic concepts, little is known about the magnitude of weathering rate fluctuations on orbital time scales. To date, proxies from marine sediments and Fe-Mn crusts that faithfully record the ocean composition over glacial interglacial cycles do not quantify the total weathering fluxes to the ocean but only indicate that the style of weathering or the source area of sediment has changed. Due to large spatial heterogeneity, individual field site measurements do not elucidate global fluxes of weathering products to the ocean and how those might affect atmospheric CO2 concentrations. Here, we use a novel approach using the pptv-level trace gas CF4, which can be analysed in air trapped in ice cores. CF4 is a trace impurity in granites and other plutonic rocks, and during weathering this gas escapes into the atmosphere. In preindustrial times, weathering of granitic rocks was the only natural source of CF4. Because CF4 is inert to destruction processes in the tropo- and stratospheres, its only sink is destruction by UV radiation in the mesosphere. This chemical inertness is responsible for an exceptionally long atmospheric lifetime which is expected to range between 50 kyr and 400 kyr. We developed a vacuum melt-extraction system for ice core samples coupled to a mass spectrometry detector to precisely measure the trace amounts of CF4 found in past atmosphere. During the last 800 kyr, the atmospheric CF4 concentrations varied in a narrow band between 31 ppt and 35 ppt, i.e. only 10-15 % variability, providing a first estimate of the long-term weathering rate fluctuations. On closer inspection, our CF4 record, however, shows a pronounced shift toward higher CF4 levels after 430 kyr (the Mid-Brunhes Event). With the beginning of Marine Isotope stage 11, we find a steep rise in CF4 that probably relates to intense weathering during the first interglacial, where CO2 reached 280 ppm and sea level may have been even higher than today. Further, our record shows that CF4 concentrations, and thus weathering, increases during interglacials and falls during the coldest, glacial phases. This dataset lends support to a strong positive coupling of continental weathering rates during warmer climate conditions at high CO2 levels.