R0021/2026-03-25/Q004/H3¶
Statement¶
Frameworks exist but were designed for traditional systems and are being extended to cover AI, with gaps and ongoing adaptation
Status¶
Current: Supported
Supporting Evidence¶
| Evidence | Summary |
|---|---|
| SRC01-E01 | FAA released AI Safety Assurance Roadmap v1 (2024) acknowledging need for new methods |
| SRC02-E01 | FDA's GMLP framework with predetermined change control plans |
| SRC03-E01 | SR 11-7 model risk management being extended to AI/ML |
Contradicting Evidence¶
No direct contradictory evidence found for H3.
Reasoning¶
All three regulated industries have validation frameworks (FAA roadmap, FDA SaMD guidance, SR 11-7), but all acknowledge that these frameworks were designed for traditional systems and require adaptation for AI. The FAA explicitly states "rigorous safety assurance methods must be developed." SR 11-7 validation approaches "may lose effectiveness" for self-adapting models.
Relationship to Other Hypotheses¶
H3 represents the nuanced reality: frameworks exist (contradicting H2) but have acknowledged gaps (qualifying H1).