The illegal addition of Sudan I dyes, synthetic azo colorants, to turmeric powder to enhance its colour poses severe health risks, including carcinogenicity. Conventional detection methods, such as high-performance liquid chromatography (HPLC) and mass spectrometry, are effective but time-consuming, costly, and require complicated sample preparation. This study presents a fast, non-destructive, and inexpensive technique of identifying Sudan I dye adulteration in Curcuma longa (turmeric) powder through Fourier Transform Infrared (FTIR) spectroscopy and sophisticated Machine Learning (ML) algorithms. Five sample groups were prepared, consisting of pure turmeric (PT) and turmeric adulterated with Sudan I dye at concentrations of 5%, 10%, 15%, and 20% (w/w), represented as T1–T4. The FTIR spectra were measured in the 400–4000 cm−1 range and were subjected to baseline correction and normalization. Principal Component Analysis (PCA) was employed to visualize spectral variations and clustering behavior among pure and adulterated samples. After that, classification models, Artificial Neural Network (ANN), Support Vector Machine (SVM), Random Forest (RF), and k-Nearest Neighbors (KNN) were trained and evaluated. Among the evaluated models, the ANN classifier achieved the highest accuracy of 99.2% in distinguishing pure and adulterated turmeric samples. These findings demonstrate the possibility of using a combination of FTIR spectroscopy and ML methods as a reliable, effective, and scalable method of detecting Sudan I dye adulteration in turmeric powder on-site to improve food safety and quality control practices.