Concepts

Data poisoning

Reference entrySafety, ethics and evaluation

Reference entryData poisoning

Data poisoning deliberately corrupts training data to damage a model, introduce hidden behavior, or manipulate its future predictions.