OpenKedge: Governing Agentic Mutation with Execution-Bound Safety and Evidence Chains

ABSTRACT

As AI agents are increasingly granted the autonomy to mutate cloud infrastructure, the risk of catastrophic, unauthorized execution grows exponentially. This paper introduces OpenKedge, a protocol and safety layer for cloud automation that validates the semantic intent of AI agents prior to execution. By binding agentic mutations to verifiable evidence chains and execution-bound safety constraints, the framework ensures that autonomous workflows operate strictly within approved operational boundaries.