IBM adds HashiCorp Vault checks to control AI agents

IBM introduced identity and control tools for AI agents to prevent them from moving money or changing payroll without verification. The package ties agent identity, policy enforcement and secret management to HashiCorp Vault, with Nvidia OpenShell named as part of the runtime layer.

• Agents get secret access only when tasks match approved rules
• The package works with IBM software and outside stacks
• Nvidia OpenShell is named as part of the policy runtime layer