Purpose and Intent as Governance Dimension in Agent-Aware IAM

This paper extends the agent-aware IAM framework (Hayagreevan, Steinacker, 2026) by introducing purpose and intent as two formally distinct, complementary governance dimensions for agentic AI. It proposes eight definitions that precisely distinguish purpose - a declared, lifecycle-governed attribute of an agent's digital identity - from intent - an inferred, session-scoped runtime signal. Two new governance constructs are introduced: the purpose path, a directed auditable record of purpose transitions over the agent's lifetime, and a two-level governance architecture connecting identity governance to runtime authorization through the containment constraint and the permission compatibility check. For multi-agent systems, a recursive containment constraint is introduced requiring both a purpose chain requirement and an intent alignment requirement to hold simultaneously. The framework maps directly to EU AI Act (Articles 9, 13, 14, 19) and GDPR (Article 5(1)(b)) obligations. Seven open research problems are identified, with the granularity problem as the most immediately consequential for implementation.