Opsin, a security company focused on governing enterprise AI agents, has secured $21m in a funding round co-led by SYN Ventures and Race Capital.

The company will put the capital towards accelerating product development, extending its reach among enterprises in regulated sectors, broadening its risk protection layer as agents gain more autonomy, and recruiting additional staff. Opsin views the investment as evidence that the market is moving towards a new security model designed specifically for agents.

According to Opsin, traditional enterprise defences were designed to watch endpoints, networks and the exchange of prompts and responses. Agents, by contrast, operate server-side, plug straight into corporate data and can take action on their own. Existing tools cannot assess whether an agent’s conduct remains consistent with the job it was built to do.

To address this, Opsin created a standardised agent model that brings consistency to how agents are classified across Microsoft 365 Copilot, Copilot Studio, Claude Enterprise, ChatGPT Enterprise and Gemini.

After identifying thousands of projects and recurring patterns of data oversharing, the firm made agent posture, covering connectors, tool calls and sharing scope, alongside runtime behaviour, the core inputs for understanding each agent. Runtime signals include an agent reaching beyond its role or passing sensitive data to users whose roles do not match.

This information feeds the Opsin Risk Graph, which learns continuously from agent activity to spot new threats. If an agent breaches policy, the platform can quarantine it, block the action or alert its owner with remediation guidance.

Source link