Opsin Raises $21M for AI agent security

On October 15, San Jose cybersecurity startup Opsin announced a $21 million financing round co-led by SYN Ventures and Race Capital, with the company positioning itself as the governance layer for enterprises deploying autonomous AI agents at scale.

Opsin was founded by James Pham, Oz Wasserman, and Jeremy Mailen — a team with deep cybersecurity roots. Pham and Wasserman previously worked together at Abnormal Security, the email-security company, before teaming up with Mailen to build Opsin around what they saw as an emerging class of risk created by AI agents operating inside corporate systems.

The company said the new capital will fund engineering recruitment, product development, and sales expansion. It did not disclose its valuation or the funding stage, according to reports on the announcement. The raise makes Opsin one of the best-funded pure-play bets on AI agent security to date.

A risk graph for the agentic enterprise

Opsin describes its platform as an AI-native governance and security system built to help enterprises adopt AI agents securely across workforce-approved AI platforms — including Microsoft Copilot, Claude, ChatGPT Enterprise, and Gemini.

According to FinSMEs, the system connects identity, agent, data, tool, behavioral, and intent context to give security teams visibility into what AI agents can access, what they are intended to do, and when their behavior deviates from that purpose. The platform is designed to discover AI agents operating across business systems, identify oversharing and excessive access, prioritize risk, and enforce security controls.

Dealroom reported that at the core of the product is a 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 of deployment.

The company already serves nearly 20 enterprise customers, including Culligan, Barry-Wehmiller, Wellstar Health System, and UiPath, spanning healthcare, financial services, manufacturing, and other regulated sectors.

The 93% problem driving AI agent security demand

Opsin's pitch lands on a striking data point from its own research: 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.

The statistic captures why AI agent security is becoming a distinct budget line rather than a feature of existing endpoint tools. Legacy security products were built around networks and devices; they were not designed to judge whether an autonomous agent's behavior still aligns with its assigned purpose. As agents move from answering questions to taking independent actions — placing orders, modifying records, touching production systems — the gap between what an agent can reach and what it should reach becomes the attack surface.

That framing was echoed by the round's investors. Race Capital general partner Alfred Chuang argued that agents are gaining access and authority over critical systems, creating a security problem enterprises have not faced before — and that the founding team's threat-detection experience at Abnormal Security made them the right team to address it. SYN Ventures partner Glenn Chisholm pointed to a fundamental trust gap around giving agents access to sensitive data, describing Opsin as the confidence layer enterprises need to embrace agents without taking on uncontained risk.

A crowded race for the agent trust layer

Opsin is not alone in the AI agent security category. The same week brought news of Shamrock AI, an Indian enterprise startup that raised $5 million in seed funding led by Endiya Partners for an AI assurance platform that tests and validates autonomous enterprise agents, according to AI Startups News. The wave of funding into AI agent security reflects a maturing view that governance must be built alongside the agents themselves, not bolted on after deployment.

The category also overlaps with identity-centric approaches: startups building agent identity and permissioning layers are attacking the same trust problem from the access-control side. RuntimeWire reported that Business Insider covered the round as an enterprise oversight play, with CEO James Pham betting that security teams will need a fundamentally different kind of monitoring as software agents embed themselves into business workflows.

The governance question travels well beyond security teams. As agents take on insurance verification, payment follow-ups, and clinical coordination in sectors like healthcare, the controls around what agents can access become central — the same challenge examined in this week's coverage of Confido Health's $12.5 million raise for AI healthcare admin agents, where patient-facing agents demand strict oversight over sensitive data.

What to watch

Opsin has not disclosed its valuation, revenue, or detailed product roadmap, which makes it difficult to assess its competitive positioning against broader cybersecurity players that may integrate similar capabilities, as AI Press Room noted. The key questions now are how quickly the company can scale past its initial customer base, and whether enterprises treat AI agent security as a standalone purchase or fold it into existing security platforms.

Either way, the $21 million round is a clear signal: as agents gain real authority inside the enterprise, the money is following the guardrails, not just the agents.