The National Institute of Standards and Technology (NIST) has released an initial public draft of NIST Special Publication (SP) 2400-102 E2026, "Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations." The publication page lists November 9, 2026 as its date, and NIST is taking public comments until December 18, 2026.
What the draft is
The draft defines terms for adversarial machine learning (AML), meaning attacks on machine learning systems and the ways to mitigate them. It sorts these attacks by the type of machine learning method, the stage of the life cycle where an attack happens, and the attacker's goals, capabilities and knowledge. It also includes a glossary of key terms, meant to give the field a common language.
NIST's pages call the subject Super Intelligence (SI) systems. According to the announcement, the new edition "focuses on attacks on Agentic SI Systems and corresponding mitigations," and it adds a separate index for agentic attacks next to the general attack index. The keywords NIST lists for the draft include data poisoning, evasion, privacy breach, large language model and chatbot.
How it relates to the earlier report
NIST says this edition builds on its AI 100-2 series. NIST AI 100-2 E2025, which the draft's page lists as a related publication, has the same title and was published as a final report in March 2025.
Who should read it
NIST describes the audience as the people and groups who design, develop, deploy, evaluate and govern these systems. Among other questions, NIST asks reviewers which new attacks threaten agentic systems and which mitigations work against them.
How to comment
The comment period runs through December 18, 2026. The publication page links the full draft as a PDF and gives an e-mail address for comments; check the address on that page before you send anything. A draft can change before the final version, so treat its terms as proposed until NIST publishes the final report.
For your career
If you work toward AI security roles and want to see how your current skills compare with a target role, try the Skill Gap Analyzer, or browse cybersecurity careers to see which roles fit you.
