IBM unveils AI software to secure and govern enterprise AI agents

IBM has launched a new software suite designed to unify AI governance and security for enterprises deploying agentic and generative AI systems at scale. Announced on 18 June 2025, the solution brings together IBM Guardium AI Security and Watson X. governance, creating a single platform for managing compliance, identifying vulnerabilities, and controlling risks tied to autonomous AI agents.

As AI agents take on more tasks in corporate workflows, IBM is focusing on the need to embed security and oversight from the outset. The new integration lets businesses detect unauthorised agents, run automated red-teaming exercises, track usage, and enforce custom security policies for both input and output data. The system currently supports compliance checks against 12 global frameworks, including the EU AI Act, ISO/IEC 42001, and New York City Local Law 144.

IBM has also partnered with AllTrue.ai to add the ability to detect AI use cases in decentralised environments, such as cloud platforms, embedded systems, and source code repositories. Once flagged, these instances can automatically trigger governance workflows through watsonx.governance, helping security teams respond faster and more consistently.

With the increased integration, watsonx.governance now monitors AI agents throughout their entire lifecycle. Enterprises can embed evaluation tools to measure performance indicators such as relevance and accuracy, and diagnose causes of underperformance. IBM plans to roll out additional features by the end of June, including agent audit trails, risk assessments, and a catalogue of approved tools.

The software also includes Compliance Accelerators, which offer pre-configured global regulatory content. This helps companies match their AI operations to standards from authorities like the U.S. Federal Reserve and the National Institute of Standards and Technology (NIST), minimising the complexity of regulatory compliance.

To support companies through the adoption process, IBM Consulting Cybersecurity Services has introduced advisory support combining IBM’s technical platforms with guidance on governance models, secure design, and policy implementation. These services will help organisations discover shadow AI deployments, assess risks, and build safeguards into every layer of their AI systems.

Industry analysts say the move responds to growing concerns about AI risks. Jennifer Glenn of IDC noted that the rapid spread of agentic AI has made it difficult for security teams to track incidents or quantify business risk. A single platform offering governance and security functions could help close that gap.

IBM has also expanded its offering geographically, making Watsonx.governance available on AWS’s data centres in India. The release includes advanced model monitoring features designed to help businesses operating under region-specific regulations.

With this launch, IBM aims to support organisations looking to scale AI safely. The unified platform is designed not only to secure data and model interactions but also to give businesses better visibility into how AI agents operate, and where things might go wrong.