Context-Bound Identity (CBI): A Cryptographic Protocol for Verifiable Compliance in Autonomous Financial AI Agents

As global capital markets transition from automated rules-based execution to Autonomous Financial AI Agents driven by Large Language Models (LLMs), traditional Identity and Access Management (IAM) frameworks face obsolescence. Standard credentials (API keys, OAuth tokens) authenticate the owner of a financial AI agent but cannot verify the state, intent, or regulatory adherence of the agent at runtime. This exposes institutions to "Agentic Risk"-where valid credentials are used by financial AI agents undergoing hallucination or prompt injection attacks to execute unauthorized high-stakes transactions. This paper introduces Context-Bound Identity (CBI), a protocol that binds authorization to a cryptographic hash of the agent's runtime instructions and risk mandates. We present the mathematical formulation for "Session-Bonded Identity" to minimize latency in High-Frequency Trading (HFT) environments and provide a comparative analysis against mTLS and OAuth 2.0.