AI Identity: Standards, Gaps, and Research Directions for AI Agents
A recent paper on arXiv (2604.23280) introduces the concept of AI Identity, which refers to the ongoing connection between the stated identity of an AI agent and its actual behavior, constrained by the reliability of their alignment. The authors perform a detailed survey of industry developments, new standards, and relevant literature, conducting a gap analysis throughout the entire lifecycle of agent identity. They provide a comparative analysis of human and AI identity across four key aspects—substrate, persistence, verifiability, and legal status—revealing significant disparities and the challenges of applying human frameworks to AI agents. The study emphasizes that AI agents are now capable of executing real transactions and workflows across organizational lines without constant human oversight, raising the unresolved issue of how to identify, verify, and hold accountable entities lacking a physical presence, enduring memory, and legal status.
Key facts
- arXiv paper 2604.23280 defines AI Identity as the relationship between declared identity and observed behavior.
- AI agents run real transactions, workflows, and sub-agent chains across organizational boundaries without continuous human supervision.
- Current infrastructure cannot identify, verify, or hold accountable AI agents lacking body, persistent memory, or legal standing.
- The paper conducts a structured survey of industry trends, emerging standards, and technical literature.
- A gap analysis is performed across the full agent identity lifecycle.
- Three contributions include a structural comparison of human and AI identity across four dimensions.
- The four dimensions are substrate, persistence, verifiability, and legal standing.
- The asymmetry between human and AI identity is fundamental, making human frameworks inadequate.
Entities
Institutions
- arXiv