The paper discusses the use of deep reinforcement learning (RL) to avoid tearing instability in tokamak plasma, a critical issue for stable and efficient fusion energy production. Tearing instability, caused by finite plasma resistivity at rational surfaces, is the leading cause of plasma disruptions. The authors developed a multimodal dynamic model to predict the likelihood of future tearing instability based on multiple diagnostic signals and actuator inputs. This model serves as a training environment for an RL-based artificial intelligence (AI) controller, which aims to maintain high plasma pressure while preventing tearing instability. The AI controller was tested in the DIII-D tokamak, the largest magnetic fusion facility in the United States, and successfully maintained low tearability under challenging conditions, including low safety factor and low torque. The controller's performance was superior to traditional preprogrammed control, demonstrating its potential for developing stable high-performance operational scenarios for future tokamak devices like ITER. The study highlights the advantages of RL over conventional control methods, including multi-actuator and multi-objective control, and the ability to consider future effects, making it a promising approach for real-time plasma control in fusion reactors.The paper discusses the use of deep reinforcement learning (RL) to avoid tearing instability in tokamak plasma, a critical issue for stable and efficient fusion energy production. Tearing instability, caused by finite plasma resistivity at rational surfaces, is the leading cause of plasma disruptions. The authors developed a multimodal dynamic model to predict the likelihood of future tearing instability based on multiple diagnostic signals and actuator inputs. This model serves as a training environment for an RL-based artificial intelligence (AI) controller, which aims to maintain high plasma pressure while preventing tearing instability. The AI controller was tested in the DIII-D tokamak, the largest magnetic fusion facility in the United States, and successfully maintained low tearability under challenging conditions, including low safety factor and low torque. The controller's performance was superior to traditional preprogrammed control, demonstrating its potential for developing stable high-performance operational scenarios for future tokamak devices like ITER. The study highlights the advantages of RL over conventional control methods, including multi-actuator and multi-objective control, and the ability to consider future effects, making it a promising approach for real-time plasma control in fusion reactors.