MITRE ATLAS 2026 Top AI Detection: Training Data Membership Inference Attack [AM
Detects adversarial attempts to infer whether specific records were used in an AI model's training set by analyzing raw prediction confidence scores and identifying suspicious query patterns such as near-duplicate submissions or systematic sweeps.
Sigma
