The persistence of organic carbon (C) in soil is most often considered at timescales ranging from tens to thousands of years, but the study of organic C in paleosols (i.e., ancient, buried soils) suggests that paleosols may have the capacity to preserve organic compounds for tens of millions of years. However, a quantitative assessment of C sources and sinks from these ancient terrestrial landscapes is complicated by additions of geologically modern (~ 10 Ka) C, primarily due to the infiltration of dissolved organic carbon. In this study, we quantified total organic C and radiocarbon activity in samples collected from 28- to 33-million-year-old paleosols that are naturally exposed as unvegetated badlands near eastern Oregon’s “Painted Hills”. We also used thermal and evolved gas analysis to examine the thermodynamic stability of different pools of C in bulk samples. The study site is part of a ~ 400-m-thick sequence of Eocene–Oligocene (45–28 Ma) paleosols, and thus we expected to find radiocarbon-free samples preserved in deep layers of the lithified, brick-like exposed outcrops. Total organic C, measured in three individual profiles spanning depth transects from the outcrop surface to a 1-m depth, ranged from 0.01 to 0.2 wt% with no clear C-concentration or age-depth profile. Ten radiocarbon dates from the same profiles reveal radiocarbon ages of ~ 11,000–30,000 years BP that unexpectedly indicate additions of potentially modern organic C. A two-endmember mixing model for radiocarbon activity suggests that modern C may compose ~ 0.5–2.4% of the total organic C pool. Thermal and evolved gas analysis showed the presence of two distinct pools of organic C, but there was no direct evidence that C compounds were associated with clay minerals. These results challenge the assumption that ancient badland landscapes are inert and “frozen in time” and instead suggest they readily interact with the modern C cycle.