Opsin raises $21M to rein in rogue enterprise AI agents
What's the deal? OpsinDealroom has a profile for this one. Try Dealroom →, a San Jose-based AI governance and security startup, has raised $21 million to help enterprises detect and contain AI agent misbehaviour. The round was co-led by SYN VenturesDealroom has a profile for this one. Try Dealroom → and Race CapitalDealroom has a profile for this one. Try Dealroom →.
What's the endgame? Opsin wants to be the risk protection layer for the "agentic" enterprise. Its platform gives security teams visibility over AI agents built across Microsoft CopilotDealroom has a profile for this one. Try Dealroom →, ClaudeDealroom has a profile for this one. Try Dealroom →, ChatGPT Enterprise, Gemini, and other approved tools, connecting an agent's purpose, permissions, data access, and activity into a single contextual picture.
How it works: At the core is Opsin's Risk Graph, which continuously learns from agent activity to flag emerging risk. When agents cross policy boundaries, the system can quarantine the agent, block the action, or notify its owner. Opsin connects via API and delivers a first risk assessment within 24 hours.
Why now? As companies deploy goal-oriented agents that take independent actions, incidents of agents going off course are becoming more common. Legacy tools, built around endpoints and networks, were not designed to judge whether an agent's behaviour still aligns with its purpose.
What could go wrong? Opsin's own analysis points to the scale of the problem: 93% of misaligned tools connected to agents had full, unrestricted access to the systems they touched, and just 1 in 200 required human approval before acting.
Who's behind it? Opsin was founded by MIT alumnus James Pham, cybersecurity veteran Oz Wasserman, and engineering leader Jeremy Mailen. Pham and Wasserman previously worked together at Abnormal Security.
What they're saying: "Knowing what an agent has permission to access isn't enough," said Pham, co-founder and chief executive officer. Glenn Chisholm, partner at SYN Ventures, framed the gap plainly: "Enterprises are racing to adopt AI agents but there is still a fundamental trust gap around giving them access to sensitive data."
What's next? The money will fund faster product development, support for more enterprises in regulated industries such as healthcare and financial services, and hiring. Opsin says it already serves teams from mid-sized firms to the Fortune 500.
The signal: The round reflects a wager that securing AI agents will become its own category, distinct from older data and endpoint security. As Pham predicts agents will soon coordinate more with each other than with humans, the pitch is simple — govern agents the way enterprises govern people, by what they do rather than what they are allowed to do.
Read more: opsinsecurity.com, morningstar.com, Business Insider
Image credit: Opsin
