Humans have indirectly influenced species at lower trophic levels by driving losses of apex consumers. Furthermore, humans have indirectly influenced species at higher trophic levels by driving losses of primary producers. Beyond these broad classes of apex consumers and primary producers, it remains challenging to identify minimum subsets of species that are particularly important for maintaining ecosystem structure and functioning. here we use a novel method at the intersection of control theory and network theory to identify a minimum set of driver node species upon which ecosystem structure strongly depends. Specifically, humans could unintentionally, completely restructure ecosystems (i.e., change species abundances from any initial values to any final values, including zero) by altering the abundances of these few critical driver node species. We then quantify the proportion of these driver nodes that arc influenced by humans, top predators, and primary producers in several marine food webs, We find that humans could unintentionally completely restructure marine food webs while only directly influencing less than one in four species. Additionally, humans directly influence: (1) most or all of the species necessary to completely restructure each network, (2) more driver nodes than top predators, and at least as many driver nodes as primary producers, and (3) an increasing proportion of driver nodes over time in the Adriatic Sea. We conclude that humans have potentially huge impacts on marine ecosystems while directly influencing only the relatively small subset of species that are currently fished. It may be possible to reduce unintentional and undesirable cascading human influences by decreasing human impacts on driver node species in these and other food webs.