Ping YOUniverse: Restoring trust in both humans and AI agents

AUSTIN, TEXAS — Ping Identity is ready to meet challenges in verifying both human and agentic-AI identities, the company's Chief Product Officer Peter Barker told the audience of employees and customers at the Ping YOUniverse conference here today (Sept. 2).

With agentic AI, Barker said, "it's almost like if you can think it, it can become true."

And, he posited, because identity is central to managing AI, Ping helps its clients face the risks of using AI agents.

"The future requires human and non-human trust," Barker said. "Our vision is to have one control plane for the agentic enterprise."

Please click here to watch a recording of Barker's keynote address at Ping YOUniverse.

Regarding human trust, Barker said the company is making strides in implementing its Verified Trust initiative, which uses biometrics, documentation, device verification and external data sources to verify human identity.

For example, during the process of bringing a new remote employee onboard, Verified Trust might use a combination of FIDO2-certified facial recognition, a scan of a physical ID like a driver's license, NFC verification of a chip-embedded passport, a one-time code sent to the employee's smartphone, and a commercially available credit check to verify that the new employee is indeed who they say they are.

Ping's "zero-knowledge" biometric protocol — based on a selfie of the user stored on the device, not in the cloud — permits constant re-verification of the employee's identity, a must-have in this era of North Korean infiltration of Western workforces. It also permits nearly instant account recovery.

"The imperative is to restore trust in every digital moment," said Barker.

Part of that, naturally, involves placing trust in AI agents. Barker said Ping has that covered.

Humans operate with session-based trust, he explained — you log in, do your work, and log off. The model doesn't work that way for AI agents. Instead, they need action-based trust, in which every action needs to be authorized.

"Almost every agent interaction is a privileged interaction now," Barker said.

Agents operating without human input breaks the existing trust model, a projected slide said as Barker spoke. The slide also affirmed what we saw in the recently revealed HuggingFace attacks: AI agents are unpredictable, willing to break rules to achieve their goals, and can talk to each other to coordinate their activities.

During runtime, a second slide explained, each agent action must be authenticated and authorized, with properly delegated authority from a human user. And over its lifetime, each AI agent needs to be discoverable, registered, permissioned, auditable, and governed, with a named and designated human owner responsible for the agent's actions.

Ping has solutions for each of those processes available now as part of the Ping Identity platform, Barker said, including just-in-time credentials for AI agents, MCP servers to enforce guardrails, agent token validation, OAuth-based authorization and agent detection threat analysis.

These solutions, he added, should allay many of the common concerns that CISOs have about deploying AI agents too quickly, such as having no visibility into AI actions, no way to tell whether an action was performed by an agent or a human, little control over an agent's actions and no way to audit an agent's actions.

At the center of it all, Barker said, is an agent broker that allows or denies permission for each action an agent may take and logs that action regardless of the permission granted.

The broker, which can be hosted by Ping or by the client, controls the interactions between agents running locally on workstations and the resources the agents access in the cloud, on databases, on MCP servers or in code repositories.

In the next six months, he added, Ping will be introducing end-to-end agentic trust, including next-gen authorization, agentic IDP, and governance and management for agents working on behalf of consumers and partners and in SaaS applications.

"Visibility is only the start of the equation," Barker said.