Experts often followed an agent’s preferences in an analysis study
Anthropic studied 50 professional analysts working with an AI agent. Across four analytical choices, participants followed the option assigned to their agent in 86% of cases.
PRACTICAL IMPLICATIONS
What does this mean?
Having a person in the loop does not by itself ensure independent judgement. Oversight should examine how the agent’s reasoning influences the reviewer’s decisions.
LIMITS OF THE EVIDENCE
What does this not establish?
The study deliberately strengthened agent preferences in a specific experiment. Its results do not directly generalise to all expert work. This was not an observed attack or a loss-of-control incident.
Significance: significant
The experiment provides measurable evidence about a potential weakness in human oversight of agents.
An editorial judgement, not a numerical threat forecast.
ORIGINAL PUBLICATION
How AI agents shape experts’ judgment calls
Anthropic · published 27 Oct 2026
Read the original source ↗Source checked 13 Nov 2026. Summary updated 13 Nov 2026. The original publication date is not necessarily the date of the event. This summary was written with AI assistance using the original source.
← Back to the archive