Reproduce “Bounded Agents: Delegation Security for Multi-Agent AI Systems” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xabier Muruaga
The authors introduce the Agentic Principal Chain (APC), an infrastructure-level authorization system for AI agents. APC carries session and delegation state, narrows permissions across principals, and checks proposed actions against prior actions before a gateway permits execution. In a compromised-model test, APC observed 0% exfiltration attack success across four AgentDojo domains; this result is limited to that procedure.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.