KnowBe4 has expanded its Agent Risk Manager to support Anthropic’s Claude, extending its AI-agent governance capabilities beyond Microsoft Copilot. The integration is designed to give organisations greater visibility into how autonomous AI agents operate and help security teams identify potentially risky activity in real time. (Business Wire)
The development comes as businesses increasingly use agentic AI for activities ranging from data analysis and workflow automation to interaction with third-party applications. According to KnowBe4’s 2026 Agentic Risk to Human Wins Report, 58% of cybersecurity leaders surveyed said AI agents are already taking actions within organisational workflows, while 52% reported AI use that was either unapproved or ungoverned. (Business Wire)
This growing use of autonomous AI creates additional security challenges because AI agents can access data, applications, APIs and organisational credentials while operating with varying degrees of human oversight.
KnowBe4 said Agent Risk Manager introduces a governance layer that monitors agent activity without requiring changes to the underlying AI model. The solution focuses on the actions, outputs and tools being used by AI agents, helping security teams detect potentially dangerous behaviour. (Business Wire)
The Claude integration provides security teams with visibility into conversations conducted through Claude.ai as well as Claude accessed through APIs. It also uses six specialised detection engines to identify risks such as prompt-injection attacks, leakage of sensitive information or personally identifiable information, privilege escalation and unauthorised access to tools. (Business Wire)
Another feature, called Blast Radius Visualization, provides an interactive network map showing APIs and credentials connected to an AI agent. This can help Security Operations teams identify systems or credentials that could be exposed if an agent is compromised.
Matt Duren, SVP of AI and Data at KnowBe4, said organisations increasingly need to secure AI agents in the same way they have traditionally focused on securing human users. The company is particularly concerned about autonomous features that may enable large language models to make decisions with limited or no direct oversight. (Business Wire)
Agent Risk Manager is currently available in early access for eligible SAT Advanced customers using direct-sale US tenant accounts. KnowBe4 said more than 70,000 organisations worldwide use its wider digital workforce security platform. (Business Wire)
The expansion reflects an emerging cybersecurity priority for insurers, banks and other financial institutions adopting generative and agentic AI. As AI systems move from providing recommendations to independently executing tasks, governance over data access, permissions, third-party connections and automated decisions is likely to become an increasingly important component of operational and cyber-risk management.
