IBM has released new governance and control systems aimed at AI agents operating inside enterprise environments, according to Seeking Alpha.
The centrepiece is HashiCorp Vault for agents, which centrally manages the credentials and confidential information that applications and agents use to reach business systems.
Vault 2.2, developed with HashiCorp, became generally available in October 2026, adding secured agent accounts, automatic credential renewal and post-quantum encryption.
The release pairs with IBM's new Agent Identity capability, which gives each agent a governed record of its owner, lifecycle status, delegated authority and actions.
It is designed to preserve authorisation context as the agent moves between systems, and to allow real-time revocation after an unauthorised attempt.
The push responds to a security gap traditional secrets management was never built for: autonomous agents that can make decisions and act across multiple systems rather than simply running as fixed applications.
IBM is layering the tools onto watsonx Orchestrate, its operational hub for discovering and governing agents, and watsonx.governance, which IBM positions as the compliance layer.
IBM cites alignment with the EU AI Act, ISO/IEC 42001 and the NIST AI Risk Management Framework, with monitoring across more than 10 languages.
IBM had already extended watsonx Orchestrate with an agentic control plane in May 2026, and watsonx.governance gained broader agent discovery and integrations with observability tools Arize and Langfuse in October 2026.
The announcement follows IBM's acquisition of data-streaming platform Confluent, completed in March, with day-one integrations into watsonx.data, IBM MQ, webMethods Hybrid Integration and IBM Z, part of a wider effort to move enterprise AI from proofs of concept to operationalised, agent-driven workflows.
Read more
