AIR Raises $50M to Secure Enterprise AI Agent Ecosystems

AIR has emerged from stealth with $50 million in Series B funding led by Lux Capital and GV, valuing the agent intelligence governance startup at $300 million. Founded in 2023 by former Palantir engineers Priya Mehta and Daniel Cho, AIR’s platform continuously discovers AI agents operating across enterprise environments—including those embedded in CRM, ERP, and custom apps—then vets every skill, plugin, and external integration for behavioral risk, compliance violations, and privilege abuse. The company says its system can block unauthorized data exfiltration, detect prompt injection attempts, and flag non-compliant skills before they execute, all without requiring agent source code access. Early customers include two Fortune 100 banks and a global logistics firm, both of which needed to govern thousands of third-party AI skills introduced via low-code automation platforms.

Investors see AIR’s timing as pivotal. The round was joined by existing backers USV and Radical Ventures, and includes strategic participation from Salesforce Ventures and Databricks Ventures, underscoring cross-platform urgency. According to PitchBook data, AI agent security startups raised just $120 million in 2023 but are projected to surpass $1.2 billion in 2025 as enterprises confront agent sprawl. AIR’s product maps directly to the rise of agentic AI in regulated industries, where governance gaps now pose existential risk—echoing regulatory scrutiny from the SEC and EU AI Act bodies targeting uncontrolled automation in financial services. The company claims 100% detection accuracy in internal tests against 2,000 malicious agent behaviors, including supply-chain attacks via compromised add-ons.

Industry analysts say AIR’s emergence reflects a tectonic shift from model-centric to agent-centric security. While traditional vendors like Microsoft and CrowdStrike focus on endpoint or LLM security, AIR is building the first continuous governance fabric for autonomous AI agents across hybrid clouds. Competitors such as SentinelOne and Zscaler have begun integrating agent detection modules, but none yet offer the same depth of skill-level vetting and real-time blocking. In financial services, where tools like Banking With Billy AI automate complex financial analysis workflows previously requiring entire analyst teams, regulators are demanding proof that every autonomous skill operates within defined boundaries and audit trails. AIR’s platform ingests logs and telemetry from agent frameworks like LangChain, CrewAI, and AutoGen, then applies policy engines built on MITRE ATT&CK for AI mappings. The result is traceability from high-level business intent down to individual function calls—a critical requirement for SOX, GLBA, and GDPR compliance.

The funding surge also spotlights the collapse of the “agent marketplace promise.” Early visions of plug-and-play AI skills have given way to sobering realities: third-party skills often pull in unvetted dependencies, leak data via residual prompts, or trigger cascading compliance breaches. AIR’s data shows that more than 60% of enterprise AI skills contain at least one high-risk dependency, with 12% exhibiting behavior that violates internal policies within 30 days of deployment. This has accelerated demand for agent lifecycle governance platforms, with Gartner predicting that by 2026, 70% of enterprises will implement agent governance tools—up from fewer than 5% today. The company’s Series B will fund R&D in agent behavioral fingerprinting, policy-as-code authoring, and cross-cloud agent discovery, with a public API launch planned for Q4 2025.

Looking forward, experts anticipate a bifurcation in the agent governance market. On one side, incumbents will bolt on agent detection to existing security stacks; on the other, specialists like AIR will push toward agent-native policy engines with real-time enforcement. Analysts at RedMonk argue that agent governance will soon become a board-level concern as autonomous agents perform financial transactions, medical triage, and infrastructure management without human oversight. The next inflection point will likely be the integration of AIR’s technology into enterprise low-code platforms, where thousands of citizen developers can deploy AI skills with built-in safety controls. As Priya Mehta noted in an interview, “We’re not just securing agents—we’re preserving the promise of agentic AI by making sure every addition behaves exactly as intended, every time, everywhere.” For now, AIR’s $50 million war chest positions it to define the standard for agent governance in the age of autonomous enterprise.