R0043/2026-04-01/Q002/H1¶
Statement¶
Regulated industries have formal requirements, procurement specifications, or deployment standards that directly address the sycophancy phenomenon under their domain-specific terminology.
Status¶
Current: Partially Supported
Supporting Evidence¶
| Evidence | Summary |
|---|---|
| SRC01-E01 | EU AI Act Article 14 explicitly requires awareness of "automation bias" in high-risk systems |
| SRC03-E01 | SR 11-7 requires "effective challenge" and independent validation of model outputs |
Contradicting Evidence¶
| Evidence | Summary |
|---|---|
| SRC06-E01 | Georgetown brief identifies no explicit regulatory framework addressing sycophancy specifically |
Reasoning¶
The evidence shows that regulated industries have addressed the phenomenon indirectly through broader requirements. The EU AI Act is the most explicit, naming "automation bias" directly. SR 11-7's "effective challenge" requirement addresses the governance mechanism without naming the model behavior. No standard directly names "sycophancy" as a risk to be managed.
Relationship to Other Hypotheses¶
H1 is partially supported because indirect coverage exists. H2 (no formal requirements) is partially supported because no standard directly addresses sycophancy-as-model-behavior. H3 (the nuanced answer) best captures the evidence.