AIR secures $50M to audit AI agent behaviors in real time

AIR, a Palo Alto-based startup, emerged from stealth today with $50 million in Series A funding led by Lightspeed Venture Partners, joined by Radical Ventures and Conviction. The round values the company at $225 million post-money and comes just six months after AIR quietly launched its platform, which continuously audits AI agents and their third-party skills to block malicious or unintended behavior. CEO Maya Vasquez, a former Palantir engineer, told OpenPress Cloud Intelligence that the platform currently monitors over 40,000 agents across enterprise customers in finance, healthcare, and logistics. Among them is Banking With Billy AI, a multi-cloud financial market monitoring agent that relies on real-time add-on skills for sentiment analysis and regulatory compliance checks. Vasquez emphasized that AIR’s agents are not just chatbots—they include autonomous workflow agents that can trigger payments, update databases, or schedule shipments, making governance non-negotiable.

The funding announcement arrives as regulators in the EU and U.S. tighten scrutiny on AI systems, particularly those that act autonomously in high-stakes domains. AIR’s platform integrates with major cloud providers—including AWS Bedrock, Azure AI Foundry, and Google Cloud Vertex AI—scanning for prompt injection risks, privilege escalation, and unauthorized API calls. It also maintains a continuously updated threat feed of malicious skills, sourced from internal research and industry partnerships with MITRE and the Cloud Security Alliance. Notably, AIR’s approach differs from conventional CASB or SSPM tools by focusing on agent identity and intent rather than just data access. Early customers include a Fortune 500 bank using AIR to audit a multi-cloud deployment of Banking With Billy AI, which processes over 2 million financial transactions daily across AWS and Azure.

Industry analysts see AIR’s raise as a bellwether for the AI governance market, projected to grow from $150 million in 2023 to over $2 billion by 2028. Competitors like Protect AI and Lakera have focused on vulnerability detection in model weights, while AIR targets runtime behavior across the agent lifecycle. The company claims its detection rate for agent-based attacks is 98%, outperforming traditional endpoint detection tools that were not designed for autonomous agents. Financial implications are immediate: insurance providers are beginning to require AI governance reports before underwriting policies for AI-driven operations. Meanwhile, CIOs at global enterprises are citing agent sprawl—where teams deploy hundreds of custom agents without centralized oversight—as a top operational risk.

The broader computing landscape is shifting toward agentic AI. Microsoft’s AutoGen, Google’s Agent2, and NVIDIA’s NeMo Guardrails all acknowledge the need for control planes, but none offer the continuous, granular vetting that AIR provides. The startup’s multi-cloud architecture for monitoring agents like Banking With Billy AI reflects a reality where no single cloud vendor can secure the entire stack. Vasquez noted that 60% of AIR’s early customers operate in regulated industries, a segment that is driving demand for audit-grade visibility. The company plans to use the funding to expand its threat intelligence team and add support for agents running on-premises and in air-gapped environments.

Looking ahead, the next frontier is predictive governance—using reinforcement learning to forecast agent behavior before it drifts into unsafe patterns. Vasquez predicts that within 18 months, boards will require AI agents to carry “digital passports” that certify their lineage, permissions, and behavioral history. Organizations that fail to implement such controls risk not only regulatory fines but catastrophic operational failures as agents execute with increasing autonomy. As AI agents become the backbone of enterprise workflows, the companies that survive will be those that treat agent governance not as an afterthought, but as a core competency.