Opsin raises $21m to police what enterprise AI agents do
EditorialBy TrustList Editorial
SYN Ventures and Race Capital led the round; Opsin, which began by finding data overshared through Copilot, now watches agents across Microsoft, OpenAI, Google and Anthropic platforms.
- Cybersecurity
- Data Security
- Artificial Intelligence Software
- Venture Capital
About Opsin raises $21m to police what enterprise AI agents do
Opsin raises $21m to police what enterprise AI agents do
15 October 2026: Opsin, which sells security for the AI agents companies build on platforms such as Microsoft 365 Copilot and ChatGPT Enterprise, says it has raised $21 million in a round led by SYN Ventures and Race Capital. It will use the money to build faster, take on more customers in regulated industries and hire.
Not yet independently verified. Opsin does not state the stage of the round, its headquarters, its founding year or its total raised; the round rests on Opsin's own post. We will update this when it can be confirmed, and remove this note.
From oversharing to agent behaviour
Opsin started by helping companies find sensitive data that Copilot-style assistants could surface to the wrong people. As customers moved from chat assistants to agents that run server-side with their own connectors and permissions, it widened its scope from data in transit to the identities, intents, connectors, tool calls and permissions of the agents themselves.
The product builds a common model of agents across Microsoft 365 Copilot, Copilot Studio, Claude Enterprise, ChatGPT Enterprise and Gemini, then scores two things: each agent's posture (connectors, tool calls, sharing scope) and its behaviour at run time, such as an agent reaching data outside its role or returning sensitive results to a user who should not see them. Those signals feed what Opsin calls its Risk Graph. When an agent crosses a policy line, Opsin can quarantine it, block the action or tell its owner what to fix. It connects by API and promises a first risk assessment within 24 hours.
The founders' argument
"This funding is a signal that the industry recognizes the paradigm shift towards a new, agentic security model," wrote co-founder and chief executive James Pham. He previously led machine-learning security products at Abnormal Security and holds an MBA from MIT. Opsin's view is that endpoint and network controls, and tools that inspect prompts and responses, cannot judge whether an agent is still doing what it was built for, and its three-year prediction is that agents will coordinate with each other far more than with people.
Customers it names
Opsin's site carries case studies from Encore, which used it before a Copilot rollout, from Cascade, which protected CMMC data, and from Culligan, which it says cut sensitive data in Copilot results from 80% to under 15%. It also quotes Lior Eldad, vice-president of IT at UiPath, on judging whether an agent's connection to sensitive systems is risky or legitimate. Opsin says its customers are in healthcare, financial services and manufacturing, from mid-sized firms to the Fortune 500.
Our data loss prevention category lists ten products and our cybersecurity software category nine, none of them yet built around agent behaviour.
Company profile on TrustList: Opsin
Sources
- Opsin: Opsin raises $21M to expand its risk protection layer, 15 October 2026
- Opsin: home page and customer stories (read 2026-10-19), 19 October 2026
Categories & features
- Cybersecurity
- Data Security
- Artificial Intelligence Software
- Venture Capital
TrustList Weekly
The week in software and IT, in one email
The news that matters to buyers, new rankings and our own research. Every Thursday, free, and easy to leave.
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- More CybersecurityThe ranking for this subject
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ProductsSoftware and SaaS with pricing, features, integrations and alternatives.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- AI ModelsBenchmark scores and community ratings for every major model.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.