IBM used TechXchange 2026 on October 27 to launch a governance stack for AI agents: Agent Identity inside HashiCorp Vault, plus new discovery, containment and evaluation tools in watsonx Orchestrate. Why it matters: enterprises are deploying agents faster than they can track who owns them, what they can access, or how to shut one down.
October 27, 2026
What happened?
At its TechXchange conference in Atlanta, IBM announced new capabilities meant to give every AI agent a verified identity, tightly scoped access, and a traceable record of its actions. The announcements span HashiCorp Vault, IBM watsonx Orchestrate, and integrations with NVIDIA OpenShell, rather than a single new model.
IBM cited Gartner research in the release saying that by 2028 the average Fortune 500 enterprise will use over 150,000 AI agents, while only 13% of organizations believe they have the right agent governance in place. Persistent credentials, unclear ownership and disconnected controls were named as the core problems.
How does Agent Identity and Vault 2.2 work?
HashiCorp Vault already manages secrets and credentials for applications and services. IBM is now extending it to agents. Agent Identity establishes the authoritative record for each agent — who owns it, its lifecycle status, and the authority it has been granted.
Paired with Vault, that identity feeds credential controls built around least-privilege access: agents get secured, short-lived credentials limited to what a task needs. If an agent tries to retrieve information or act outside its approved permissions, IBM says access can be denied and the session revoked in real time. Vault 2.2, now generally available, gives agents their own secured accounts, automatically renews their credentials, and adds post-quantum encryption. A new Secure Hub, in public preview, is meant to give teams centralized visibility across their Vault environments.
What can watsonx Orchestrate now discover, contain and evaluate?
Orchestrate is positioned as a cross-platform control point, not a single-vendor console. A Discover capability, in tech preview, connects to platforms such as Amazon Bedrock, Microsoft Foundry and Google Gemini to find agents built elsewhere, along with models and MCP servers. A Contain capability, also in tech preview, runs agents in isolated sandboxes and gives teams a kill switch to shut down any agent. The Evaluate tools let teams test quality, accuracy and compliance before deployment, monitor agents in production, and track model and token costs.
IBM also said Agent Identity and Vault integrate with NVIDIA OpenShell, pairing a policy-governed runtime with verified identity and short-lived credentials so each action can be traced, limited and reviewed.
Why it matters
This is the unglamorous layer that will decide whether agents move from pilots to real work. A helpful agent with a long-lived, overprivileged credential is a security incident waiting for a trigger — and recent agent breaches have shown that is not hypothetical. IBM’s bet is that identity, revocation and evaluation become a paid, separate line item in the enterprise AI stack, sold across clouds rather than inside one vendor’s agent platform. For buyers, the practical test is availability: Vault 2.2, Agent Identity and Orchestrate’s evaluate capabilities are generally available now, while discovery and containment — arguably the most useful parts — are still in tech preview.
Frequently asked questions
Is IBM Agent Identity available now?
Yes. IBM lists HashiCorp Vault 2.2, Agent Identity and watsonx Orchestrate’s evaluate capabilities as generally available. Secure Hub is in public preview, and Orchestrate’s discover and contain capabilities are in tech preview.
Does this only work with IBM agents?
No, according to IBM. Orchestrate’s discovery is designed to find agents built on Amazon Bedrock, Microsoft Foundry and Google Gemini, as well as models and MCP servers.
What problem does a kill switch solve?
IBM says containment runs agents in isolated sandboxes so a rogue agent cannot interfere with others, and teams can shut down any agent directly — a response to agents acting outside their intended task.
Sources: IBM via PR Newswire (TechXchange 2026 announcement, October 27, 2026)