KnowBe4 announced Agent Risk Manager support for Anthropic’s Claude, extending its agent governance layer to Claude environments with existing native support for Microsoft Copilot. According to the KnowBe4 2026 Agentic Risk to Human Wins Report, 58% of cybersecurity leaders state that AI agents are taking actions within organisational workflows, while 52% of respondents reported their use of AI is unapproved or ungoverned. The integration provides real-time visibility and automated threat detection to govern autonomous AI agents, monitoring agent execution without modifying the underlying model. Matt Duren, SVP of AI and data at KnowBe4, said it is important to protect what agents are doing, especially when employees use features like skip permissions mode, which allows the LLM to make its own decisions autonomously with zero oversight. Agent Risk Manager offers one-time setup for insight into conversations via Claude.ai or by API, real-time threat interception with six specialised detection engines analysing prompt injections, sensitive data/PII leaks, privilege escalations, and unapproved tool access, and blast radius visualisation through an interactive Tool Network map displaying connected APIs and credentials. The product is in early access for SAT Advanced customers on direct-sale, US-tenant accounts.