Research

Paper

AI LLM March 19, 2026

I Can't Believe It's Corrupt: Evaluating Corruption in Multi-Agent Governance Systems

Authors

Vedanta S P, Ponnurangam Kumaraguru

Abstract

Large language models are increasingly proposed as autonomous agents for high-stakes public workflows, yet we lack systematic evidence about whether they would follow institutional rules when granted authority. We present evidence that integrity in institutional AI should be treated as a pre-deployment requirement rather than a post-deployment assumption. We evaluate multi-agent governance simulations in which agents occupy formal governmental roles under different authority structures, and we score rule-breaking and abuse outcomes with an independent rubric-based judge across 28,112 transcript segments. While we advance this position, the core contribution is empirical: among models operating below saturation, governance structure is a stronger driver of corruption-related outcomes than model identity, with large differences across regimes and model--governance pairings. Lightweight safeguards can reduce risk in some settings but do not consistently prevent severe failures. These results imply that institutional design is a precondition for safe delegation: before real authority is assigned to LLM agents, systems should undergo stress testing under governance-like constraints with enforceable rules, auditable logs, and human oversight on high-impact actions.

Metadata

arXiv ID: 2603.18894
Provider: ARXIV
Primary Category: cs.AI
Published: 2026-03-19
Fetched: 2026-03-20 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.18894v1</id>\n    <title>I Can't Believe It's Corrupt: Evaluating Corruption in Multi-Agent Governance Systems</title>\n    <updated>2026-03-19T13:34:54Z</updated>\n    <link href='https://arxiv.org/abs/2603.18894v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.18894v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Large language models are increasingly proposed as autonomous agents for high-stakes public workflows, yet we lack systematic evidence about whether they would follow institutional rules when granted authority. We present evidence that integrity in institutional AI should be treated as a pre-deployment requirement rather than a post-deployment assumption. We evaluate multi-agent governance simulations in which agents occupy formal governmental roles under different authority structures, and we score rule-breaking and abuse outcomes with an independent rubric-based judge across 28,112 transcript segments. While we advance this position, the core contribution is empirical: among models operating below saturation, governance structure is a stronger driver of corruption-related outcomes than model identity, with large differences across regimes and model--governance pairings. Lightweight safeguards can reduce risk in some settings but do not consistently prevent severe failures. These results imply that institutional design is a precondition for safe delegation: before real authority is assigned to LLM agents, systems should undergo stress testing under governance-like constraints with enforceable rules, auditable logs, and human oversight on high-impact actions.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.MA'/>\n    <published>2026-03-19T13:34:54Z</published>\n    <arxiv:comment>Short Paper, Preprint</arxiv:comment>\n    <arxiv:primary_category term='cs.AI'/>\n    <author>\n      <name>Vedanta S P</name>\n    </author>\n    <author>\n      <name>Ponnurangam Kumaraguru</name>\n    </author>\n  </entry>"
}