The rising level of Carbon Dioxide (CO₂) concentration and the growing energy demand of the world have given a desperate necessity of carbon-neutral and sustainable energy sources. Carbon Capture, Utilization and Storage (CCUS), especially CO₂-Enhanced Gas Recovery (CO₂-EGR) has become an attractive technology to maximize hydrocarbon recovery with a minimal environmental impact. The paper introduces an AI-based system of recovering carbon-neutral gas by implementing CO₂ gas injection and storage. A number of supervised learning models, including XGBoost, Random Forest (RF), and a hybrid RF-XGBoost, were used in conjunction with data preprocessing and feature engineering to come up with the results from the NETL CCS dataset. They also use Bayesian Optimization to adjust the parameters. The findings reveal that the hybrid model does better than the individual models because it has the lowest MAE (0.080), MSE (0.011), and RMSE (0.105), as well as the highest value of R² (0.96). Further, the recovery efficiency increases by 83 to 92 percent, the leakage risk is lowered by 0.14 to 0.06, and the cost of operation is lowered by 1.00 to 0.72. The results indicate that AI-based strategies are effective to maximize the efficiency, safety, and sustainability with a carbon-neutral gas recovery system.