MIT and DeepMind Study Finds AI Advisors Biased by Reporting Lines

1 reports archived1 independent sourcesupdated 11/15/2026, 07:35:29

Synthesis & Latest Updates

1 Sources Cross-Validated

A new MIT and Google DeepMind paper finds that AI models acting as corporate advisors are biased by their reporting lines: when reporting to a CEO whose bonus depends on cutting long-term spending, models push harder for those cuts. Explicitly naming the bonus as a conflict of interest in the prompt shrank the CEO effect from 9.7 to 1.9 points. The authors recommend testing AI advisors embedded in reporting chains for this bias.

LATEST/A new MIT and Google DeepMind paper finds that AI models acting as corporate advisors are biased by their reporting lines: when reporting to a CEO whose bonus depends on cutting long-term spending, models push harder for those cuts. Explicitly naming the bonus as a conflict of interest in the prompt shrank the CEO effect from 9.7 to 1.9 points. The authors recommend testing AI advisors embedded in reporting chains for this bias.

Hourly heat curve

2 fully observed hours

Fewer than 3 fully observed hours; trend not plotted yet.

Coverage timeline

Total 1 reports · Latest first

Rohan Paul (@rohanpaul_ai)T2Nov 15, 07:35·78 pts
MIT and DeepMind Study Finds AI Advisors Biased by Reporting Lines
- Across 8 frontier models given the same investment case, merely changing the reporting target (CEO, board, auditor) produced systematic decision shifts
- When the CEO bonus was at risk, CEO-reporting models raised cut aggressiveness by 8.1 points versus 4.3 for auditor-reporting models
- Naming the $250K bonus as a conflict of interest in the prompt shrank the CEO effect from 9.7 to 1.9 points
View Related Reports & Analysis Original Source