Skip to content

R0043/2026-03-28/Q002/SRC05/E01

Research R0043 — Sycophancy Vocabulary
Run 2026-03-28
Query Q002
Source SRC05
Evidence SRC05-E01
Type Factual

314 risk categories across 8 government and 16 corporate policies; overreliance risks 'less frequently specified in detail' in corporate policies

URL: https://arxiv.org/html/2406.17864v1

Extract

314 risk categories across 8 government and 16 corporate policies; overreliance risks 'less frequently specified in detail' in corporate policies

Relevance to Hypotheses

Hypothesis Relationship Strength
H1 Contradicts Requirements exist but are indirect/general, not direct anti-sycophancy provisions
H2 Contradicts Requirements addressing the phenomenon do exist
H3 Supports Requirements are general governance/trustworthiness rather than specific behavioral constraints

Context

Part of the pattern identified across Q002: regulated industries address AI behavioral risks through general governance frameworks rather than specific system-behavior requirements.