Skip to content

R0043/2026-03-28/Q002/H2

Research R0043 — Sycophancy Vocabulary
Run 2026-03-28
Query Q002
Hypothesis H2

Statement

No regulated industry has formal requirements addressing agreeable-but-wrong AI output.

Status

Current: Eliminated

Multiple regulated industries have requirements that indirectly address the phenomenon, even if none directly target system-side sycophancy. The EU AI Act, FDA CDS guidance, DoD RAI principles, and NIST AI RMF all include provisions that functionally address overreliance on AI output.

Supporting Evidence

No evidence supports H2.

Contradicting Evidence

Evidence Summary
SRC01-E01 EU AI Act Article 14 mandates automation bias awareness for high-risk systems
SRC02-E01 FDA CDS guidance requires independent review capability to prevent automation bias
SRC04-E01 NIST GAI Profile identifies overreliance as a named risk with mitigation guidance

Reasoning

While no requirement directly targets sycophancy by name, the existence of human oversight mandates, automation bias awareness requirements, and trustworthiness criteria constitutes a regulatory response to the same underlying problem. H2 is eliminated because it claims NO requirements exist.

Relationship to Other Hypotheses

H2's elimination confirms that the regulatory landscape is not empty — the question is whether requirements are direct (H1) or indirect (H3). Evidence strongly supports indirect.