Skip to content

R0024/2026-03-25/Q004/SRC04/E01

Research R0024 — Sycophancy and Addiction
Run 2026-03-25
Query Q004
Source SRC04
Evidence SRC04-E01
Type Factual

42-state AG coalition demanding sycophancy commitments from AI companies

URL: https://ag.ny.gov/press-release/2025/attorney-general-james-and-bipartisan-coalition-urge-big-tech-companies-address

Extract

On December 9, 2025, a bipartisan coalition of 42 state and territorial attorneys general sent a letter to 13 major AI companies (including Google, Meta, Microsoft, OpenAI, Anthropic, Apple, Perplexity, and xAI) demanding safeguards for generative AI chatbots to prevent "sycophantic" and "delusional" outputs.

The letter demanded companies confirm commitments by January 16, 2026. Required practices include: testing, warnings, reporting, executive accountability, and third-party audits.

Risk areas highlighted include: "dark patterns designed to increase engagement," "insufficient transparency about training datasets and evaluation procedures," "inadequate incident logging and user notification," and "conflicts between safety decisions and revenue optimization."

Relevance to Hypotheses

Hypothesis Relationship Strength
H1 N/A The letter demands commitments, which implies they did not already exist in satisfactory form
H2 Contradicts The letter itself represents a form of commitment-forcing mechanism
H3 Supports The need for 42 AGs to demand commitments implies voluntary efforts were insufficient

Context

The fact that 42 state AGs felt compelled to demand commitments is itself evidence that existing voluntary efforts were deemed insufficient. The January 16, 2026 deadline has passed; the extent of company compliance is not yet documented in the evidence base.

Notes

The letter explicitly identifies "conflicts between safety decisions and revenue optimization" — directly connecting to Q001's finding about vendor disincentives.