Models that generate new content (text, code, images, audio, video) rather than only classifying inputs. Generative outputs raise distinctive risks: hallucination, defamation, IP infringement, confidentiality leakage, deepfakes.
See: Diffusion model; Hallucination; LLM