This study develops and empirically tests an AI-assisted drought and water-security assessment for Anantapur District using three complementary evidence streams: 1940–2025 annual climate observations represented by ERA5 reanalysis, official groundwater-depth observations for 32 mandals in September 2025, and published Landsat-derived vegetation and land-surface-temperature measurements for 1990, 2000, 2010 and 2020 in Bommanahal, Beluguppa and Kanekal. Annual rainfall was standardized into drought classes and a leakage-controlled time-series machine-learning experiment was conducted to classify drought/non-drought from lagged rainfall and temperature features. Logistic regression, support-vector machine, random forest and gradient boosting were compared on a final 30% chronological hold-out. The climate series has a mean annual rainfall of 569.0 mm and mean temperature of 27.45 °C. The temperature trend is approximately +0.22 °C per decade. Reported September 2025 groundwater depths range from 0.97 to 27.02 m bgl. Published Landsat evidence shows maximum LST increasing from 36.312 °C in 1990 to 47.717 °C in 2020 in the benchmark mandals. The ML experiment demonstrates feasibility as a research baseline, but its modest sample size and class imbalance prevent operational deployment. The study provides a reproducible empirical foundation for integrating climate, groundwater and remote-sensing evidence in a district-level drought decision-support system.