Abstract We investigate interseismic locking on the Kamchatka megathrust, source of the 2025 Mw 8.8 earthquake, using a data‐driven probabilistic framework. Horizontal GNSS velocities constrain a boundary element model on a triangulated megathrust geometry, while an algorithmic sampling strategy estimates the locking probability for fault elements without prescribing asperity locations a priori. The resulting probability distribution reveals a locked region whose cumulative slip deficit since the previous megathrust earthquake (Mw 9.0, 1952) yields a moment deficit comparable to the 2025 release. Hierarchical clustering of the locking models identifies smaller asperities that coincide with the 2025 rupture initiation zone. Slip in the 1952 and 2025 events shows strong spatial correspondence with high locking probabilities, indicating that persistent asperities dominate moment release. Lower shallow slip in 2025 contrasts with the greater near‐trench slip inferred for the 1952 tsunami, highlighting variable shallow slip as a major source of uncertainty for tsunami hazard estimation.