Opsin's platform covers agents that employees build in Microsoft Copilot, Claude, ChatGPT Enterprise, Gemini and other enterprise-approved AI tools. It maps each agent's purpose, permissions, data access, tools and connected systems, and compares the agent's activity against its intended purpose to flag meaningful deviations. Its website lists modules for AI agent discovery, agent runtime security, AI agent governance, MCP governance, agent identity security, AI risk assessment and ongoing oversharing protection.[1],[3]
In its own analysis of agent activity across enterprise customers, Opsin found that 93% of misaligned tools connected to agents had full, unrestricted access to the systems they were connected to, and that just 1 in 200 required human approval before taking action. Pham told Business Insider that at one firm, an agent built by a chief financial officer inadvertently gave other employees top-level access to financial information. He compared agents to people: "you don't understand what is going on in a human head, but you can judge them by their actions."[1],[2]
Opsin says it serves nearly 20 enterprise customers, including Culligan, Barry-Wehmiller, Wellstar Health System and UiPath, in healthcare, financial services, manufacturing and other regulated sectors. On its website, Opsin says Culligan cut sensitive data in Copilot results from 80% to under 15% using its software. These are company-reported figures.[1],[3],[2]