Model-to-Model Negotiation Protocols in Secure Cloud Environments

The rapid evolution of cloud-native artificial intelligence systems has led to the deployment of multiple autonomous machine learning models operating concurrently within shared infrastructure. These models increasingly require direct coordination to negotiate resources, trust boundaries, execution contexts, and data-sharing constraints. Traditional centralized orchestration and policy enforcement mechanisms struggle to scale under such dynamic and decentralized conditions. This paper proposes Model-to-Model Negotiation Protocols (M2M-NP), a secure and decentralized framework that enables autonomous models to negotiate operational and security parameters in zero-trust cloud environments. The proposed architecture integrates cryptographic identity, policyas-code enforcement, and verifiable negotiation workflows. Experimental evaluation on a Kubernetes-based hybrid cloud testbed demonstrates reduced coordination latency, improved resource utilization, and stronger isolation guarantees compared to centralized orchestration approaches.