Capsule Security debuts with $7 million funding to secure AI agent behavior

Capsule Security announced its launch and published research exposing critical prompt-injection vulnerabilities (ShareLeak in Microsoft Copilot Studio—CVE-2026-21520—and PipeLeak in Salesforce Agentforce) that can allow AI agents to be manipulated to perform unsafe actions or exfiltrate data; Capsule also introduced ClawGuard and a runtime protection platform to detect and block unsafe agent behavior.