A malicious type of cyberattack in which data used for analytics, machine learning, or artificial intelligence systems is intentionally manipulated, corrupted, or misleadingly labeled in order to distort system outputs, degrade performance, or introduce bias. Data poisoning can occur during data collection, curation, integration, or model retraining. It may be difficult to detect once embedded in complex data pipelines.

As governments and institutions increasingly rely on data-driven tools for planning and oversight, data poisoning has become a critical concern for data governance, transparency, and public accountability. In governmental, educational, and workforce policy contexts, data poisoning poses significant risks by undermining evidence-based decision-making, skewing labor market intelligence, misinforming funding or regulatory actions, and eroding trust in AI-supported public systems.