On the Fragility of AI Agent Collusion First version: September 25, 2025

Jussi Keppo, Yuze Li, Gerry Tsoukalas, Nuo Yuan · 2026

Read On the Fragility of AI Agent Collusion First version: September 25, 2025 by Jussi Keppo, Yuze Li, Gerry Tsoukalas, Nuo Yuan (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-20281.txt.

Description

Keppo and co-authors test whether AI pricing agents really collude as earlier work suggested. In their model, differences in patience or data access shrink the set of collusive outcomes. Tests with open-source language agents matched those predictions. Unequal patience cut price lift from 22% to 10% above competitive levels. Unequal data access cut it to 7%. More rival agents and mixed algorithm types also broke collusion up. Only size gaps held collusion together, through leader-follower patterns.

Catalog record

CREATORJussi Keppo, Yuze Li, Gerry Tsoukalas, Nuo Yuan
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE93131
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.20281

Browse the full library · About the Tower of Babel