An integrity problem is a security issue where information is changed in a misleading way without becoming obviously false. The content may be reordered, emphasized, omitted, softened, or framed so that users draw a different conclusion than they would from a neutral presentation. In cyber security, integrity means preserving the intended meaning, structure, and trustworthiness of data and outputs.
This matters in AI systems, search, messaging, and election tools because attackers do not always need to fabricate facts; they can manipulate salience and framing. A chatbot may mention one option first, hide another, or give unequal detail, creating subtle persuasion while appearing accurate. Defenses include auditing output ordering, checking retrieval and ranking layers, logging prompts and responses, and testing for systematic bias or prompt injection. In practice, integrity controls aim to ensure that users receive information as intended, not just factually plausible text.