Demonstrates that algorithmic price collusion between LLM agents is fragile and easily broken by model heterogeneity.
March 24, 2026
Original Paper
On the Fragility of AI Agent Collusion
arXiv · 2603.20281
The Takeaway
Challenges the regulatory fear of autonomous price fixing by showing that real-world factors (diverse model sizes, data access, and patience) naturally destabilize collusive equilibria, suggesting that 'algorithmic diversity' is a valid antitrust defense.
From the abstract
Recent work shows that pricing with symmetric LLM agents leads to algorithmic collusion. We show that collusion is fragile under the heterogeneity typical of real deployments. In a stylized repeated-pricing model, heterogeneity in patience or data access reduces the set of collusive equilibria. Experiments with open-source LLM agents (totaling over 2,000 compute hours) align with these predictions: patience heterogeneity reduces price lift from 22% to 10% above competitive levels; asymmetric dat