Saved in:
Bibliographic Details
Main Authors: Summerfield, Christopher, Luettgau, Lennart, Dubois, Magda, Kirk, Hannah Rose, Hackenburg, Kobi, Fist, Catherine, Slama, Katarina, Ding, Nicola, Anselmetti, Rebecca, Strait, Andrew, Giulianelli, Mario, Ududec, Cozmin
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2507.03409
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909675765628928
author Summerfield, Christopher
Luettgau, Lennart
Dubois, Magda
Kirk, Hannah Rose
Hackenburg, Kobi
Fist, Catherine
Slama, Katarina
Ding, Nicola
Anselmetti, Rebecca
Strait, Andrew
Giulianelli, Mario
Ududec, Cozmin
author_facet Summerfield, Christopher
Luettgau, Lennart
Dubois, Magda
Kirk, Hannah Rose
Hackenburg, Kobi
Fist, Catherine
Slama, Katarina
Ding, Nicola
Anselmetti, Rebecca
Strait, Andrew
Giulianelli, Mario
Ududec, Cozmin
contents We examine recent research that asks whether current AI systems may be developing a capacity for "scheming" (covertly and strategically pursuing misaligned goals). We compare current research practices in this field to those adopted in the 1970s to test whether non-human primates could master natural language. We argue that there are lessons to be learned from that historical research endeavour, which was characterised by an overattribution of human traits to other agents, an excessive reliance on anecdote and descriptive analysis, and a failure to articulate a strong theoretical framework for the research. We recommend that research into AI scheming actively seeks to avoid these pitfalls. We outline some concrete steps that can be taken for this research programme to advance in a productive and scientifically rigorous fashion.
format Preprint
id arxiv_https___arxiv_org_abs_2507_03409
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Lessons from a Chimp: AI "Scheming" and the Quest for Ape Language
Summerfield, Christopher
Luettgau, Lennart
Dubois, Magda
Kirk, Hannah Rose
Hackenburg, Kobi
Fist, Catherine
Slama, Katarina
Ding, Nicola
Anselmetti, Rebecca
Strait, Andrew
Giulianelli, Mario
Ududec, Cozmin
Artificial Intelligence
We examine recent research that asks whether current AI systems may be developing a capacity for "scheming" (covertly and strategically pursuing misaligned goals). We compare current research practices in this field to those adopted in the 1970s to test whether non-human primates could master natural language. We argue that there are lessons to be learned from that historical research endeavour, which was characterised by an overattribution of human traits to other agents, an excessive reliance on anecdote and descriptive analysis, and a failure to articulate a strong theoretical framework for the research. We recommend that research into AI scheming actively seeks to avoid these pitfalls. We outline some concrete steps that can be taken for this research programme to advance in a productive and scientifically rigorous fashion.
title Lessons from a Chimp: AI "Scheming" and the Quest for Ape Language
topic Artificial Intelligence
url https://arxiv.org/abs/2507.03409