R0021/2026-03-25/Q004 — ACH Matrix¶
Matrix¶
| H1: Rigorous requirements | H2: Minimal/vague | H3: Adapting, incomplete | |
|---|---|---|---|
| SRC01-E01: FAA roadmap, "methods must be developed" | + | -- | ++ |
| SRC02-E01: FDA GMLP 10 principles + change control | ++ | -- | + |
| SRC03-E01: SR 11-7 "may lose effectiveness" for adaptive AI | + | -- | ++ |
| SRC04-E01: NIST voluntary RMF with TEVV | + | -- | + |
Legend: ++ Strongly supports · + Supports · -- Strongly contradicts · - Contradicts · N/A Not applicable
Diagnosticity Analysis¶
Most Diagnostic Evidence¶
| Evidence ID | Why Diagnostic |
|---|---|
| SRC03-E01 | SR 11-7's explicit acknowledgment that validation approaches "may lose effectiveness" for AI is highly diagnostic — it confirms frameworks exist (contradicting H2) while revealing their limitations (supporting H3 over H1). |
Least Diagnostic Evidence¶
| Evidence ID | Why Non-Diagnostic |
|---|---|
| SRC04-E01 | NIST's voluntary framework is non-diagnostic because its voluntary nature makes it equally consistent with H1 (framework exists) and H3 (incomplete). |
Outcome¶
Hypothesis supported: H3 — Frameworks exist but are adapting, with acknowledged gaps for modern AI systems.
Hypotheses eliminated: H2 — Multiple regulatory frameworks are in active use.
Hypotheses inconclusive: H1 — Partially supported; frameworks are rigorous but not yet fully adapted to AI.