Proof-of-Continuity: New model prevents authority escalation in AI agents

Nicola Gallo's paper "Proof-of-Continuity: A Temporal Model for Authority Propagation in Distributed Systems and AI Agents" tackles a fundamental flaw in current authorization models. Traditional Proof-of-Possession approaches derive authority from artifacts like tokens or credentials, but they fail to preserve the causal relationship between a request's origin and the authority exercised at later steps. This is especially critical for AI agents that invoke multiple tools and downstream services in a chain. The proposed Proof-of-Continuity model introduces Proof of Relationship, a single-hop causal primitive whose transitive composition ensures each execution step is causally linked to the previous one and may only propagate a non-expansive subset of the original authority. Under this model, the confused deputy condition—where a downstream service is tricked into using authority it shouldn't have—cannot be satisfied as valid behavior. Authority sources are carried together along the chain but are never merged into a combined authority; each step is authorized only against the context of the lineage that caused it. The paper focuses on authorization propagation after authentication (using OIDC, verifiable credentials, etc.) is established. This has direct relevance to distributed systems and AI agent architectures where executors hold multiple authority sources across service boundaries. By enforcing temporal causality, Proof-of-Continuity provides a minimal discipline to prevent authority escalation without replacing existing authentication mechanisms.