Catastrophic regime shifts in ecosystems occur when the system is tipped into a new attractor state under some external forcing. Here we consider whether evolutionary adaptations within ecosystems can trigger similar transitions. We use an individual-based, evolutionary model of interconnected ecosystems to analyze nonlinear changes in global state resulting from local adaptations. Transitions between periods of stability occur when new traits arise that allow exploitation of underutilized resources. Subsequent rapid growth of the population carrying the new trait causes abrupt environmental change that drives incumbent species extinct. We call these transitions 'evolutionary regime shifts'. These internally generated perturbations can result in ecosystem collapse, followed by recovery to an alternate stable state, or occasionally system-wide extinction. While these disruptions may have a negative impact on ecosystem productivity in individual simulation runs, mean results over many simulations show a trend for increasing ecosystem productivity and stability over time. Feedback between life and the abiotic environment in the model creates a 'long-tailed' distribution of extinction sizes without any external trigger for large extinction events.