Security / AI Governance / Data Quality
NIST opens comment on draft adversarial machine learning taxonomy focused on agentic system attacks
The initial public draft of SP 2400-102e2026 builds on the NIST AI 100-2 series and adds an agentic attacks index. Comments are due by 18 December 2026.

NIST published an initial public draft of Special Publication 2400-102e2026, Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations, on 9 November 2026. The edition builds on the earlier NIST AI 100-2 series and aims to give a common language that can inform future standards and practice guides on system security.
This edition centres on attacks against agentic systems and their mitigations, set out in Section 4, with a new agentic attacks taxonomy index. NIST’s page now refers to these as ‘Super Intelligence (SI)’ systems. According to CDO Magazine, the draft covers untrusted retrieved data, malicious tools and skills, multi-agent workflows, long-running tasks and memory poisoning. It also lists defences such as separating trusted instructions from untrusted content, least-privilege access, sandboxing and audit logging.
The document is voluntary. Comments are due by 18 December 2026.
Why it matters
Several of the attacks described, including data poisoning, indirect prompt injection through retrieved content and memory poisoning, enter through the data that agents read rather than through the model itself. Data owners should therefore know which knowledge bases, files and memory stores their agents can retrieve from and who is able to write to them. The taxonomy's terminology is a useful shared vocabulary for data, security and AI teams, both when they write agent risk registers and when they respond to the consultation.
Status: Consultation. Initial public draft; comments open until 18 December 2026
Key dates
- Public comment period on NIST SP 2400-102e2026 draft closes
Sources
Was this useful?
Thanks. What was the problem?
Thank you — this helps us choose what to cover. If you spotted an error, please tell us what it is.
Prepared with AI assistance and reviewed by the editor before publication. Spotted an error? Tell us — see our editorial policy and methodology.
Governance areas
Institutions
Who should care
- CISO
- AI governance lead
- CDO
- Data stewards