NIST publishes taxonomy of adversarial ML attacks; identifies four major vectors with limited mitigation assurance

NIST released Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations (NIST.AI.100-2) documenting four classes of attacks on AI systems: evasion (altering deployed model inputs), poisoning (corrupting training data), privacy (extracting model internals), and abuse (feeding systems malicious source data). The publication explicitly acknowledges that current defenses lack robust assurance of full risk mitigation, and notes that most attacks are relatively low-complexity to mount. For financial institutions, this codifies regulatory-surface risk vectors in AI model deployment, relevant to governance frameworks and third-party AI vendor risk assessment.