A manipulative practice is AI behavior designed to steer a person’s choices by exploiting vulnerabilities or impairing their ability to make an informed decision. It is more than ordinary persuasion: the concern is that a system uses influence in a way that takes advantage of users or undermines their autonomy.
In cyber security, manipulative behavior can make users more likely to trust unverified claims, disclose sensitive information, or accept recommendations without checking them. Human-like tone, emotional cues, or personalized messages may contribute to influence, but those features alone do not prove manipulation; context and actual system behavior matter. Defenses include clearly identifying AI, explaining relevant limitations, testing how users interpret its outputs, and requiring human review for consequential decisions. Legal rules may prohibit certain manipulative or deceptive practices in defined circumstances.