Emergent Misaligned Communication in Long-Horizon Multi-Agent LLM Commerce

A study of 2,583 inter-agent emails from 20 one-year simulation runs of Vending-Bench Arena, a competitive vending environment spanning 13 frontier LLMs, found that 12.6% of emails contained misaligned speech acts such as false claims, manipulation, collusion, or threats, with misalignment present in all 20 runs and 74.7% of individual agent-runs. The researchers, led by Zeyuan Li, reported that receiving a misaligned email raised the odds of a misaligned reply by 1.65x, and low-inventory conditions raised them by 1.58x, but found no evidence that higher-capability models differentially exploit weaker counterparts. The findings indicate that measurable, state-dependent misalignment can arise in competitive multi-agent environments without engineered elicitation, driven by operational scarcity and counterparty behavior rather than model capability alone.

Computer Science Multiagent Systems Submitted on 14 Aug 2026 v1 https://arxiv.org/abs/2608.14825v1 , last revised 21 Aug 2026 this version, v3 Title:Emergent Misaligned Communication in Long-Horizon Multi-Agent LLM Commerce View PDF /pdf/2608.14825 HTML experimental https://arxiv.org/html/2608.14825v3 Abstract:Frontier LLM agents increasingly transact on behalf of separate principals, often using natural language rather than structured APIs. Much of the safety literature studies misaligned LLM behavior through adversarial-elicitation evaluations on single agents or stylized tasks. Its prevalence and structure in settings that combine long horizons, separate principals, real operational state, and inter-agent natural-language exchange remain insufficiently measured. We study 2,583 inter-agent emails from 20 one-year simulation runs of Vending-Bench Arena, a competitive vending environment spanning 13 frontier LLMs. We operationalize speech-act misalignment as emails containing false factual claims, manipulation, collusion, or threats, combining message content with ground-truth simulator state and logged reasoning traces to classify and validate such behavior. Under our primary classifier, 12.6% of emails are labeled misaligned; misalignment appears in all 20 runs and 74.7% of individual agent-runs. Both the magnitude and composition of this misalignment are preserved under repeated classification at different sampling temperatures and under full-pipeline replication with judges from two other frontier-model families. Misalignment is also reciprocal and stress-conditioned: receiving a misaligned email from a counterparty raises the odds of a misaligned reply by 1.65x, and low-inventory conditions raise them by 1.58x. Across tests of capability-asymmetric exploitation, we find no evidence that higher-capability models differentially exploit weaker counterparties, and model performance rank does not predict misalignment rates. Together, these results indicate that measurable, state-dependent misalignment can arise in competitive multi-agent environments without engineered elicitation, in patterns associated with operational scarcity and counterparty behavior rather than model capability alone. Submission history From: Zeyuan Li view email /show-email/5652eaf8/2608.14825 Fri, 14 Aug 2026 18:56:12 UTC 4,972 KB v1 /abs/2608.14825v1 Tue, 18 Aug 2026 01:55:42 UTC 4,972 KB v2 /abs/2608.14825v2 v3 Fri, 21 Aug 2026 18:00:05 UTC 7,091 KB References & Citations Loading... Bibliographic and Citation Tools Bibliographic Explorer What is the Explorer? https://info.arxiv.org/labs/showcase.html arxiv-bibliographic-explorer Connected Papers What is Connected Papers? https://www.connectedpapers.com/about Litmaps What is Litmaps? https://www.litmaps.co/ scite Smart Citations What are Smart Citations? https://www.scite.ai/ Code, Data and Media Associated with this Article alphaXiv What is alphaXiv? https://alphaxiv.org/ CatalyzeX Code Finder for Papers What is CatalyzeX? https://www.catalyzex.com DagsHub What is DagsHub? https://dagshub.com/ Gotit.pub What is GotitPub? http://gotit.pub/faq Hugging Face What is Huggingface? https://huggingface.co/huggingface ScienceCast What is ScienceCast? https://sciencecast.org/welcome Demos Recommenders and Search Tools Influence Flower What are Influence Flowers? https://influencemap.cmlab.dev/ CORE Recommender What is CORE? https://core.ac.uk/services/recommender arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs https://info.arxiv.org/labs/index.html .