Comprehensive survey of machine unlearning covering definitions, scenarios, verification methods, and applications. Cited in the International AI Safety Report 2025 as a pioneering paradigm for removing sensitive information.