Feralization of genetically engineered (GE) crops increases the risk that transgenes will become integrated into natural and naturalizing plant populations. A key assumption of the management of GE crops is that populations of escaped plants are short-lived and therefore the risks they pose are limited. However, few populations of escaped crop plants have been tracked over the long term so our understanding of their persistence in ruderal or natural landscapes is limited. We repeated a large-scale road survey of feral GE canola populations in North Dakota, USA, initially conducted in 2010. Our objectives in 2021 were to determine the current distribution of feral canola populations, and to establish the relative frequency of GE and non-GE phenotypes in populations of canola throughout North Dakota. Our results indicate that, although the incidence of feral canola was less in 2021 than 2010, escaped canola populations remain common throughout the state. The prevalence of alternate forms of GE herbicide resistance changed between surveys, and we found an overabundance of non-GE plants compared to the frequency of non-transgenic forms in cultivation. Indirect evidence of persistence includes sampling plants with multiple transgenic traits, and finding populations far from transportation routes. We conclude that feral canola populations expressing transgenic herbicide resistance are established outside of cultivation, that they may be under selection for loss of the transgene, but that they nonetheless pose long-term risks by harboring transgenes in the unmanaged landscape.