Abstract: Stability monitoring of the open-pit slope is the last key line of defense for ensuring the safety of mine personnel and integrity of production equipment. Setting scientific, accurate, and reliable early warning thresholds can enhance monitoring effectiveness. However, the current open-pit slope monitoring system in China is lacking in scientific theoretical support and methodology in setting early warning thresholds. To address this problem, this study proposes an innovative method for constructing early warning thresholds based on slope creep theory. The core idea of this method is to deeply explore the internal development law of the slope creep curve, especially the evolution characteristics of cumulative displacement with time. The research focuses on the three classic stages of the slope creep process (initial, steady state, and accelerated). The displacement ratio, determined by an analysis of the critical point between the three stages, can be used as a key scientific basis in determining and grading warning thresholds. First, the representative rock mass mechanical parameters of the target open-pit mine C were selected, and a simplified rock mass model, which can reflect the key structural characteristics, was constructed using three-dimensional (3D) particle flow code (PFC) discrete element numerical simulation software. The stress–strain curves of rock mass under different stress levels were obtained via a numerical simulation of a conventional triaxial compression test. Based on the simulated stress–strain curves and creep theory, the key parameters of the constitutive equation describing the creep behavior of rock mass—namely, the elastic modulus and viscosity coefficient—were determined. Additionally, the concept of the locking segment failure mechanism was introduced, and renormalization group theory was applied to quantify the influence of the spatial variability of rock mass on macro strength and deformation. The creep proportional relationship was combined with locked segment theory and the quantitative results of spatial variability to derive a critical displacement threshold, which is more in line with the actual geological conditions. Based on the evolution characteristics of the critical displacement threshold and creep curve, thresholds for a four-level early warning system were determined to provide a hierarchical response basis for different stages of deformation. To verify the accuracy and practicability of the proposed method, the derived early warning thresholds were compared with the long-term cumulative displacement data obtained by the mine site monitoring global navigation satellite system (GNSS). The results show that the threshold constructed based on creep theory can effectively capture the key deformation stage before slope instability. Taking open-pit mine C as an example, the creep theory early warning based on cumulative displacement sent an alarm when the Level I early warning threshold (6216 mm) was reached, 11.6 days ahead of the final landslide occurrence time, providing a valuable time window for emergency response and personnel and equipment evacuation. At the short-term acceleration stage near the landslide, the landslide occurred only 3 h after the Level I warning threshold (97.2 mm) triggered by the deformation rate (tangent angle criterion), capturing the rapid acceleration process before the final instability. The accuracy and reliability of the early warning system were improved through the synergy of long- and short-term early warning mechanisms.