NIST SP 2400-102e2026 (Initial Public Draft): 'Adversarial Machine Learning — A Taxonomy and Terminology of Attacks and Mitigations'

On 9 November 2026 (confirmed via meta article:published_time and CSRC date), NIST released the initial public draft of Special Publication 2400-102e2026, 'Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations.' It builds on the historical NIST AI 100-2 series and establishes a taxonomy of AML attacks and mitigations plus standardized terminology for the field of adversarial machine learning. This edition expands the scope to attacks on Agentic SI (Super Intelligence) Systems and their mitigations (Section 4), and updates the 'SI Attacks Taxonomy Index' with a new 'Agentic SI Attacks Taxonomy Index' for fine-grained referencing. Authors include Apostol Vassilev (NIST), Alina Oprea (Northeastern), Alie Fordyce (ElevenLabs), Xander Davies (UK AI Security Institute), Maia Hamin (NIST). Public comment period is open through 18 December 2026.

This is the authoritative common-language reference for adversarial ML attacks on agentic/SI systems, intended to inform other standards and future practice guides for assessing and managing SI system security. It directly extends the AI 100-2 lineage that underpins NIST AI RMF implementation and is expected to be referenced by enterprises deploying agentic AI, red teams, and conformity-assessment programs. The agentic-SI focus aligns with the current wave of autonomous-agent security risk.

Review the draft and submit comments by 18 December 2026, particularly on the new Agentic SI attacks taxonomy, mitigation effectiveness, and terminology requiring standardization. Map your own SI/agentic-AI threat models and control catalogues to the taxonomy.