NIST AI 200-2 ipd: a draft framework for testing AI systems, not yet a final publication

What we observe

NIST's page announces "Input Sought on Initial Public Draft of NIST AI 200-2 through October 6, 2026", with an announcement date of 7 August 2026 [1]. The document is titled The TEVV-Athlon Framework for Evaluating AI Systems, carries the identifier "NIST AI 200-2 ipd (Initial Public Draft)" and is dated August 2026 [2]. TEVV stands for test, evaluation, verification and validation [1, 2].

The suffix matters. NIST's own rules for publication identifiers state that the most common public comment document is the Public Draft, "and specifically the Initial (first) Public Draft (IPD)", and that "The initial PD has <stage> = ipd" [3]. After reviewing comments on a public draft, authors decide whether to issue a further public draft or, more commonly, a final publication; later drafts are labelled 2pd, 3pd and so on, or fpd for a final public draft, and a final publication carries no stage label [3]. The draft itself says: "NIST is releasing this document for public comment" [2]. Its publication history reads: "Approved by the NIST Editorial Review Board on [Will add date upon final publication]" [2]. NIST AI 200-2 ipd is therefore a draft for comment, not NIST's final guidance.

The draft states: "A 60-day comment period opened August 7, 2026, and closes on October 6, 2026", with comments to be emailed with "NIST AI 200-2" in the subject line [1, 2]. NIST invited input on six areas, including definitions of the TEVV terms, the framework's flexibility and scope, activities it may not address, and aspects that "may warrant clarification, revision, removal, or expansion" [1].

In substance, the draft introduces "a four-stage method for developing customized assessments of AI systems based on organizational TEVV objectives", in which AI systems "are tested via a set of Events and Tools which produce data on Blocks related to measurement concepts of interest" [1, 2]. The four stages are Articulate & Organize, Define & Construct, Apply & Measure, and Synthesize & Interrogate [2]. NIST presents the framework as complementing the core functions of its AI Risk Management Framework: Govern and Map feed the first stage, the TEVV-Athlon "itself implements the Measure core function", and results from the fourth stage inform Manage [2]. NIST says it is intended to cover statistical machine learning models, large language models, multi-modal models, agentic systems and other AI technologies [1].

Status on 28 October 2026. NIST's page still describes the document as the initial public draft [1]. NIST's technical publications list of drafts includes the entry "The TEVV-Athlon Framework for Evaluating AI Systems (Draft)", NIST AI 200-2 ipd [4]. A request on 28 October 2026 for the PDF at the address a final NIST AI 200-2 would use returned "not found" (HTTP 404), and the addresses a second, third or final public draft would use also returned 404 [5]. We found no final version or second draft. The comment period has closed, so a revised version may follow; NIST has not announced a date.

Why it matters

NIST links this work to its AI Risk Management Framework, which, the draft notes, "specifically calls for a Test, Evaluation, Verification, and Validation (TEVV)" methodology [1, 2]. A shared method for designing assessments could make results from different evaluators easier to compare. But a draft can change after comments are reviewed [3]. Any citation of NIST AI 200-2 should keep the "ipd" label, should not describe the TEVV-Athlon Framework as NIST's finished guidance, and should be checked against NIST's page for a final version.

Method

Sources: NIST's announcement page, the draft PDF, NIST's publication identifier syntax document (April 2022) and NIST's list of draft technical publications, all saved and read on 28 October 2026 [1-4]. Status was checked by requesting the addresses where a final version and a second, third or final public draft would be published under NIST's identifier rules [3, 5]; a 404 shows only that no document was at that address on that date. Quotations keep NIST's US spelling. The draft is a NIST publication with named NIST authors; we cite it by title and identifier.

References

1. [1] National Institute of Standards and Technology. "The TEVV-Athlon Framework for Evaluating AI Systems" (announcement, 7 August 2026). https://www.nist.gov/artificial-intelligence/ai-research/tevv-athlon-framework-evaluating-ai-systems. Accessed 28 October 2026.
2. [2] National Institute of Standards and Technology. NIST AI 200-2 ipd (Initial Public Draft), The TEVV-Athlon Framework for Evaluating AI Systems, August 2026. https://doi.org/10.6028/NIST.AI.200-2.ipd (https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.200-2.ipd.pdf). Accessed 28 October 2026.
3. [3] National Institute of Standards and Technology, Information Services Office. Publication Identifier Syntax for NIST Technical Series Publications, April 2022. https://www.nist.gov/system/files/documents/2022/04/01/PubID_Syntax_NIST_TechPubs.pdf. Accessed 28 October 2026.
4. [4] National Institute of Standards and Technology. NIST Technical Publications List, drafts. https://pages.nist.gov/NIST-Tech-Pubs/DRAFT.html. Accessed 28 October 2026.
5. [5] NIST Library publications server, requested 28 October 2026: https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.200-2.pdf (HTTP 404), https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.200-2.2pd.pdf (HTTP 404), https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.200-2.3pd.pdf (HTTP 404) and https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.200-2.fpd.pdf (HTTP 404).