Maximum fluid-induced earthquake magnitude shifts from injected volume to stress control

Fluid injection during subsurface operations can trigger earthquakes. Traditional hazard frameworks estimate maximum magnitude from injected volume, assuming self-arrested ruptures within the fluid-pressurized zone. Unexpected large induced earthquakes exceed these predictions, suggesting rupture propagation is not always bounded by the pressurized zone. Recent numerical and laboratory studies suggest elevated background stress may promote runaway rupture, but the transition between regimes and its implications for maximum magnitude remain unresolved. Using a large-scale experimental apparatus and complementary finite-element simulations, we identify two rupture regimes governed by distinct maximum-magnitude scaling laws. Self-arrested ruptures remain bounded by the fluid-pressurized zone, with maximum magnitude controlled primarily by injected volume. We derive equivalent operational and mechanistic scalings for this regime. Runaway ruptures propagate beyond the fluid-pressurized zone, with maximum magnitude controlled by effective background stress rather than injected volume. We directly resolve the transition between regimes, providing the first evidence for the theoretical critical prestress threshold of Garagash and Germanovich. Together, these results reconcile volume-based hazard models with the occurrence of large induced earthquakes and establish a framework linking rupture regime to the maximum magnitude of fluid-induced earthquakes.