AI Agent Identity Becomes the Next Security Challenge
17m
AI Agent Identity Moves Into the Spotlight
AI agent identity is becoming a critical issue as enterprises start to imagine agents handling sensitive tasks on behalf of people. In this Techstrong TV interview, Mike Vizard talks with Philipp Pointner, chief of digital identity at Jumio, about how organizations should think about trust, delegation and accountability in an agentic world.
Pointner compares an AI agent to an assistant running an errand for a CEO. The assistant must prove who they are, show what the CEO asked them to do and sometimes prove the CEO’s identity as well. That same model applies to digital agents. Teams need to know which human an agent belongs to, what the agent may do and whether the transaction requires human proof.
Delegation and Traceability Matter
The conversation explores the emerging standards around verifiable credentials, delegation protocols and agent permissions. Pointner notes that many building blocks already exist, but practical implementations are still limited. Most websites and services do not yet support agents that need clear signposts, scoped access and auditable interactions.
AI agent identity also raises hard questions about traceability. When organizations receive an identity proof, they must know its source. Did it come directly from the person or from the person’s agent? That distinction will matter for banks, payments, onboarding and any workflow where identity fraud creates risk.
Fraudsters Are Already Using AI
Pointner warns that attackers are adopting generative AI and agentic tools quickly. Fraud operations that once required teams of people can now use bots, scripts and adaptive models to test defenses repeatedly. That makes identity verification more important, not less.
Reusable digital identities can help reduce risk because cryptographic proof stays with the user. Instead of copying and reusing passport scans or face images, businesses can receive verified identity data while the proof mechanism remains under the user’s control.
Start Small, But Do Not Stand Still
The interview also addresses regulation, payments and the need for more agentic security systems. Pointner expects new attack vectors and believes defenders will need AI to counter AI-driven threats in real time.
For security and technology leaders, the practical advice is to experiment carefully. Start with low-risk use cases. Learn how protocols such as MCP might change digital interactions. Give developers room to understand the technology. The worst option is to do nothing, because AI agent identity will become a foundational part of securing the next generation of digital services.