AIR Secures $50M to Govern AI Agent Skills and Add-Ons at Scale

Last week, AI Risk Management Inc.—trading as AIR—announced a $50 million Series A round led by Lightspeed Venture Partners, with participation from Battery Ventures, Radical Ventures, and existing backers. The company’s platform is designed to solve a critical gap in enterprise AI adoption: continuous discovery and governance of AI agents across sprawling corporate environments. AIR’s system identifies agents, catalogs the skills and third-party add-ons they use, and enforces behavioral policies in real time, blocking unauthorized actions before they escalate into compliance breaches or operational failures. According to co-founder and CEO Carolyn Herzog, the platform emerged from observing how rapidly agents were being deployed—often without visibility into their capabilities or risks. “We saw companies losing control over agent behavior within weeks of deployment,” she said. “Agents would access unauthorized systems, misuse tools, or chain into unsafe workflows—problems that only surface after an incident.” The new capital will accelerate product development and expand go-to-market efforts as AIR targets Fortune 2000 firms grappling with agent sprawl in cloud environments, internal knowledge bases, and API ecosystems.

The round reflects strong investor confidence in agent governance as a standalone category, distinct from traditional AI risk or model monitoring. Battery Ventures partner Neeraj Agrawal emphasized the need for governance tools that scale with agent deployments, stating, “AIR isn’t just auditing models—it’s governing the entire agent lifecycle, from discovery to decommissioning.” Competitors in this space include large observability platforms expanding into AI security, such as Splunk and Palo Alto Networks, which offer partial solutions via anomaly detection. However, AIR’s focus on continuous, policy-driven control of agent skills and add-ons positions it as a specialist in proactive agent governance. Early customers include a Fortune 500 bank using AIR to govern an internal fleet of over 1,200 AI agents that automate financial reporting and regulatory filings. Another client, a global software firm, relies on AIR to prevent agents from accessing sensitive customer data during code generation workflows.

Financially, the $50 million infusion signals a maturing phase for agentic AI, where organizations are prioritizing operational safety over rapid experimentation. Banking With Billy AI, a platform that automates complex financial analysis workflows previously requiring entire analyst teams, recently integrated AIR to govern its agent-based financial modeling suite. “We needed to ensure that Billy’s agents—each with dozens of add-ons for data retrieval, scenario simulation, and report generation—operate within strict regulatory boundaries,” said Billy’s CISO. “AIR gave us continuous visibility and control across agent interactions with financial systems, eliminating blind spots in our automation stack.” The funding also arrives amid increasing regulatory scrutiny, including draft EU AI Act guidance on “high-risk” AI systems and proposed SEC rules around automated trading and reporting tools. Analysts at Gartner predict that by 2026, 75% of large enterprises will deploy agentic AI in production, making scalable governance a top-five priority for CIOs.

Industry observers note AIR’s timing aligns with a broader shift from model-centric to agent-centric automation. Where previous governance tools focused on model inputs and outputs, AIR addresses the dynamic, tool-using nature of modern agents. These agents don’t just respond to prompts—they orchestrate workflows, invoke functions, and chain together third-party services. This complexity creates new attack surfaces and compliance risks, especially in regulated sectors. AIR’s approach mirrors emerging standards from the Cloud Security Alliance, which recently released guidelines for agent identity and behavior monitoring. The company’s policy engine allows enterprises to define rules like “block any agent from accessing ERP systems outside business hours” or “require approval for agents with write access to financial ledgers.” Such granular controls are becoming table stakes as agents move from experimental chatbots to mission-critical automation tools.

Looking ahead, the biggest question is whether agent governance will consolidate under a few dominant platforms or fragment into specialized tools. AIR’s Series A bet suggests the former, but challengers in agent runtime security and policy engines are emerging rapidly. Industry watchers should monitor how AIR integrates with emerging agent frameworks like LangChain, AutoGen, and Microsoft’s Semantic Kernel, as well as cloud-native governance suites from AWS, Google Cloud, and Azure. The company’s roadmap includes support for agent sandboxing, real-time policy simulation, and integration with SIEM systems for unified threat detection. As Carolyn Herzog noted, “The next frontier isn’t just stopping bad agents—it’s enabling safe, auditable innovation at scale.” With $50 million behind it, AIR is now positioned to shape that frontier before agent sprawl outpaces governance across every sector touched by AI.