From Saying to Doing: Opsin Tames the New Breed of Autonomous AI Agents

SAN FRANCISCO, CA – June 17, 2026 – The enterprise is at a critical inflection point. After years of experimentation, artificial intelligence is moving out of the lab and into production, with AI-powered agents poised to redefine workforce productivity. But a significant hurdle remains: security. As these agents evolve from simple copilots that “say” things to autonomous actors that can “do” things, they introduce a new class of risks that legacy security tools were never designed to handle. Into this high-stakes environment steps Opsin, a two-year-old security firm that claims to have cracked the code for governing this new, agentic workforce.

The company has reported significant momentum in 2026, citing major customer wins in highly regulated industries, key analyst recognitions, and a series of platform innovations. Opsin argues it represents a “second generation” of AI security, moving beyond static prompt analysis to provide real-time governance for agents acting on their own across an organization's entire tech stack.

A New Security Paradigm: From Chatbots to Agents

The fundamental challenge, according to cybersecurity experts, is that autonomous agents behave in ways that are dynamic and non-deterministic. Unlike traditional software with predictable logic, an AI agent can connect to new data sources, chain together complex decisions, and execute actions at machine speed, often without direct human oversight. This creates a vast and poorly understood attack surface.

“Legacy tools were built for documents at rest and humans at keyboards. The first waves of AI security solutions were built for prompts and chatbots. Neither was built for what's actually happening now: agents acting on their own across your stack,” said James Pham, Co-Founder and CEO of Opsin, in a recent announcement. “Security teams cannot govern that with rules written for the last decade.”

This is where Opsin claims its core differentiation lies. The platform is built around a contextual layer, powered by its ‘AI Context Graph,’ which launched in March 2026. This technology maps the complex relationships between AI agents, the data they access, and the human identities they act on behalf of. The goal is to answer the critical questions that keep CISOs up at night: What is this agent doing? What sensitive data is it touching? Why is it doing that, and who is ultimately responsible?

This approach has earned the company recognition from industry watchers. In 2026, Opsin was included as a Representative Vendor in a Gartner® Market Guide for Guardian Agents and featured in an OWASP GenAI Security Landscape report. This external validation suggests that the market is beginning to recognize the unique security challenges posed by agentic AI and the need for specialized solutions.

Measurable Impact in Regulated Industries

For enterprises in sectors like healthcare and manufacturing, the risks of AI failure are not theoretical; they carry material operational, regulatory, and reputational consequences. Opsin’s most compelling claims center on its tangible results within these high-stakes environments, where it is now in production at organizations like Wellstar Health System, Barry-Wehmiller, and Culligan.

The company’s “1-24 promise”—one-click API onboarding and a full risk assessment within 24 hours—is designed for rapid time-to-value. At global manufacturing firm Barry-Wehmiller, this initial assessment found that over 70% of AI prompts referenced overshared data. After deploying Opsin, the firm reduced that exposure to less than 20%, giving it the confidence to expand its Microsoft Copilot usage. Similarly, water treatment company Culligan saw sensitive data appearing in Copilot queries drop from 80% to under 15% after implementing the platform’s remediation guidance.

These individual wins are part of a broader pattern of reported outcomes across Opsin’s customer base. The company claims its platform has led to a 70% average reduction in AI risk within six weeks and a 67% reduction in the time it takes to trace a problematic AI interaction back to its root cause—from over 30 minutes of manual log digging to under 10 minutes of automated analysis. Perhaps most tellingly, Opsin reports that its customers see an average 253% month-over-month growth in enterprise AI adoption once the security team gives the green light, suggesting that effective governance is the primary enabler for scaling AI initiatives.

“CISOs at regulated enterprises know that agentic AI is moving faster than their existing security stack,” noted Opsin Co-Founder and CPO Oz Wasserman. “They need a partner who has purpose built a solution for what comes next. That’s the role Opsin is playing.”

Building the Governance Layer for a Multi-AI World

Few enterprises are standardizing on a single AI platform. The reality is a complex ecosystem of tools from Microsoft, Google, Anthropic, and others, creating fragmented security visibility. Opsin is positioning itself as a unified governance layer that sits above this ecosystem. The company launched on the Microsoft Azure Marketplace in March 2026, simplifying procurement for joint customers, and has built integrations for Microsoft Copilot, Copilot Studio, ChatGPT Enterprise, Google’s Gemini, and Anthropic’s Claude.

This broad support is critical. While native tools like Microsoft Purview provide essential capabilities, they can create a “Microsoft-only view of risk.” By offering a single pane of glass across a mixed-agent estate, Opsin aims to provide a comprehensive inventory and risk posture that prevents security gaps between platforms.

A key innovation is the platform's support for Claude’s managed agents—autonomous entities that can perform background tasks on a user's behalf. This capability directly addresses the shift from human-in-the-loop copilots to fully autonomous workflows, a frontier where governance and control are paramount.

A Platform Built for the Agentic Future

Opsin organizes its platform capabilities across three core pillars designed to address the full lifecycle of AI security. The first, Data Oversharing & Posture Management, continuously scans for exposed sensitive data and decentralizes remediation by sending automated instructions to business data owners. The second, Detection & Response, provides behavioral monitoring of live AI interactions to catch active misuse, such as jailbreak attempts or unintentional data exfiltration. The third and most forward-looking pillar, Agentic AI Governance, provides a full inventory of all custom agents, classifies their intent, and monitors for behavioral drift, with the ability to instantly quarantine risky agents.

Beyond these pillars, the company continues to ship new features like the Opsin Agent Defense, an on-demand assessment of agent-related security risks. This continuous innovation cycle is focused on a single trajectory: securing the future of work.

“We started by solving data oversharing in Copilot. We are now securing an entire generation of autonomous enterprise agents,” Pham added. “The enterprises that win will be the ones whose security teams can confidently say ‘yes, deploy it’ because Opsin gives them the visibility, context, and controls to govern it.”