IBM Tackles the Agent Governance Gap at TechXchange

New tools aim to help enterprises track, secure and control fast-growing rosters of AI agents

AI agents are proliferating, promising to do everything from booking flights, to balancing inventories, to building software. But for these new digital workers to truly be effective at scale, they need to be governed. 

IBM made that challenge a central focus of TechXchange 2026 last week, with several product announcements aimed at helping enterprises manage their burgeoning rosters of AI agents. These agents can be spun up in seconds and operate at machine speed. The tools introduced by IBM are meant to address those challenges as enterprises put more and more agents to work. 

By 2028, the average Fortune 500 enterprise will be deploying more than 150,000 agents, according to Gartner. There is still much progress to be made, however, to ensure that adequate controls are in place to truly leverage these tools. Only 13% of those enterprises think they have sufficient governance in place to manage the agents that are expected to materialize.

But the will appears to be there as enterprises see the value of infusing their operations with tireless, autonomous digital agents. “Autonomous brings scale. That’s why enterprises are moving there,” Dinesh Nirmal, senior vice president of software at IBM, said at TechXchange. 

IBM is introducing the following new products as organizations seek that scale, striving to make Gartner’s prediction come true.

Agent Identity

Introduced at TechXchange, Agent Identity acts as a centralized registry cataloging agents in the enterprise. Agent Identity assigns each agent a cryptographically verifiable identity, tracking key information including who created the agent, its lifecycle status and its authority level. 

“Every agent should be registered, should be owned and should be sponsored. And that is an imperative,” Tyler Lynch, outbound CTO of IBM Automation, said in a press briefing introducing the new products ahead of TechXchange.

In addition to serving as a registry, Agent Identity allows human users to take action as needed to govern and isolate agents. To have agents act on their behalf, human users can issue OAuth 2.0 On-Behalf-Of (OBO) tokens through Agent Identity. 

These tokens are time-limited and tightly scoped to specific API endpoints or resource IDs. As Lynch explained, an OBO “allows a human to delegate to an agent with very, very fine-grained access controls about what that agent can do on that person’s behalf and for how long they can do it.”

Tracking agents, permissions and intent will help solve a major challenge in agent deployment. During a conversation on the TechXchange stage, Drew Roy, director of product management at HashiCorp and IBM Verify, outlined that challenge: “Who is this agent? Who is it acting on behalf of? What does it want to do and what does it actually do when it accesses end systems? And that information will give us the assurance that any policies or governance controls we put in place and allow us to push forward.”

HashiCorp Vault 2.2

While Agent Identity works at the identity layer, HashiCorp Vault 2.2 works in concert at the credentials layer. Vault 2.2 issues agents short-lived and least-privilege credentials, also performing auto-renewals.

“The role that Vault and Agent Identity are playing in this whole thing is trying to ensure that the agent operating inside of the sandbox never has persistent access to the credential it needs to access any systems,” Roy said. “ … And so it’s really reducing that overall blast radius and providing auditability and sort of traceability as the agent behaves inside the ecosystem.”

Vault 2.2 also advances quantum-safe security with post-quantum cryptography built into its public key infrastructure (PKI) capabilities. To prepare for future quantum computers that could compromise today’s widely used public-key cryptography, the secrets management platform now supports quantum-resistant digital signatures for X.509 certificates.

Additionally, Vault 2.2 acknowledges the “secret zero” problem for supported on-premises workloads: how an agent, application or system proves its identity when being issued its first credential. To handle the problem, Vault 2.2 anchors identity to physical hardware instead of a static password or software key. 

Secure Hub

Sitting above Agent Identity and Vault in the agent security stack is another product announced at TechXchange, IBM Secure Hub, a centrally managed visibility layer that allows for the monitoring, auditing and managing of all Vault clusters across the infrastructure. 

“IBM Secure Hub governs at scale, bringing centralized visibility and reporting to your credentials and secrets with operational visibility, no matter where your Vault clusters are on the cloud, on-prem, or at edge,” Lynch said. 

In deploying agents at enterprise scale, launching them is the easy part. Tracking each one of these digital identities is the hard part, Lynch said. That challenge is compounded by the fact that non-technical knowledge workers will be creating these agents. 

The solution, he said, is visibility: “You can’t govern what you can’t see.”

---