🎯Capsule Security is like a security guard for AI helpers in businesses. They just got $7 million to help keep these AI helpers safe from bad actions and mistakes, and they have some really smart people advising them on how to do it.
The Development
Capsule Security, an Israeli startup focused on securing AI agents, has officially launched with $7 million in seed funding led by Lama Partners and Forgepoint Capital International. Founded in 2025 by Naor Paz and Lidan Hazout, the company aims to prevent AI agents from being manipulated, misbehaving, or exfiltrating sensitive data during workflows. The platform continuously monitors AI agents' behavior, providing visibility and intervention capabilities to correct anomalous or unsafe activities in real-time.
Security Implications
The startup's solution operates as an independent layer compatible with various frameworks, ensuring that it can monitor agents’ reasoning, interactions, and executions without disrupting their operations. Capsule’s platform supports several models, including Cursor, Claude Code, Microsoft Copilot Studio, ServiceNow, and Salesforce Agentforce, and can be integrated seamlessly into existing workflows.
Notably, Capsule Security is advised by prominent figures in the cybersecurity field, including Chris Krebs, the first Director of CISA, and Jim Routh, former CISO for multiple Fortune 500 companies. Krebs emphasized the critical nature of monitoring AI agents, stating, "Legacy tools weren’t built to monitor what happens between prompt and action—that’s the runtime gap. Capsule closes it."
Industry Impact
The rapid adoption of AI agents in enterprise environments has raised concerns, with Microsoft reporting that over 80% of Fortune 500 companies are now utilizing active AI agents. This surge underscores the need for robust security measures to protect against potential threats that arise from AI's operational speed and complexity.
Capsule Security has also published two significant research reports identifying zero-day vulnerabilities in major platforms: ShareLeak, a critical severity vulnerability in Microsoft Copilot Studio, and PipeLeak, a vulnerability in Salesforce Agentforce. These vulnerabilities demonstrate how untrusted inputs can influence agent behavior, posing significant risks to enterprise operations.
What to Watch
As AI agents continue to evolve, Capsule Security's innovative approach aims to secure this new class of privileged users. The company has developed ClawGuard, an open-source enforcer that adds a pre-invocation checkpoint before agents execute tool calls, enhancing security measures across various platforms. With the increasing integration of AI agents into business systems, the need for real-time visibility and control over agent behavior is more crucial than ever.
Capsule's technology not only blocks unsafe or unauthorized activities but also generates auditable telemetry designed for governance and compliance teams, ensuring that organizations can adopt AI agents confidently while maintaining security and trust.
The rapid integration of AI agents into enterprise environments necessitates robust security solutions like those offered by Capsule Security, which address the unique challenges posed by these advanced technologies.
