AI models differ in openness depending on whether companies share access, weights, or training data, creating closed, open-weight, and open-source categories. Organizations customize models through system prompts, retrieval-augmented generation, fine-tuning, or full training, weighing convenience against control, transparency, and performance.