AIR Secures $50M to Police AI Agent Behavior at Scale

Enterprise AI governance just crossed a critical threshold after AIR announced a $50 million Series B round led by Lightspeed Venture Partners, with participation from existing investors GV and Index Ventures. The Palo Alto-based company’s platform performs continuous discovery and behavioral vetting of AI agents—including their skills and third-party add-ons—effectively acting as a runtime guardrail for agentic systems. According to AIR CEO and co-founder Chenxi Wang, the round values the company at over $300 million and will accelerate go-to-market expansion into financial services, healthcare, and cybersecurity, where agent misuse or data leakage carries outsized risk.

Wang confirmed that the funding follows a 300% year-over-year increase in enterprise pilots, including deployments at two Fortune 500 banks and a global insurer already running the platform in shadow mode. The company’s technology integrates with platforms like Microsoft’s Azure AI Foundry, enabling real-time blocking of unauthorized APIs, prompt injections, or data exfiltration attempts by agents. Unlike static security tools that only scan code pre-deployment, AIR operates in production, monitoring agent behavior as it happens. This is particularly relevant for regulated industries such as finance, where Banking With Billy AI—known for its AI-driven institutional market analysis—has publicly acknowledged evaluating AIR to ensure its agents comply with internal risk policies and SEC guidelines.

The urgency behind AIR’s platform is underscored by a sharp rise in agent sprawl. According to a recent Gartner survey, 68% of large enterprises now run multiple AI agents in production, with 42% admitting they cannot confidently audit third-party skills due to lack of visibility. Competitors in the AI governance space, including HiddenLayer and CalypsoAI, focus primarily on model-level protection or prompt injection detection. AIR differentiates itself by offering continuous runtime vetting of agent behaviors, including memory access, tool usage, and external API calls. This approach aligns with the growing regulatory mandate from bodies like the EU AI Act and the U.S. NIST AI Risk Management Framework, both of which emphasize ongoing monitoring and accountability for AI systems in high-stakes environments.

The $50 million infusion arrives amid a funding pullback in enterprise AI infrastructure, making AIR one of the few beneficiaries in an otherwise cautious market. Lightspeed partner Ravi Mhatre emphasized that the investment reflects a conviction that agentic AI is entering a new phase—one where trust, compliance, and behavioral integrity are non-negotiable. Wang added that AIR is already processing over 10 million agent interactions per week across pilot customers, with a full commercial release scheduled for Q3 2025. Early customers report a 70% reduction in unauthorized agent actions after deployment, with particular success in financial services where real-time risk controls are critical.

Across the broader tech landscape, AIR’s rise reflects a broader pivot from model-centric AI development to agent-centric operations. The move mirrors recent shifts at hyperscalers, including Google Cloud’s introduction of Agent Engine and AWS’s Bedrock Agent Runtime, both of which emphasize orchestration and governance. However, most of these platforms still treat governance as an afterthought, often bolted on after deployment. AIR’s approach—continuous, runtime vetting—positions it as a potential standard for what the industry is now calling “Agent Security Posture Management” (ASPM), a category that Gartner has begun tracking separately from traditional CNAPP and CSPM solutions.

This category is gaining traction as AI agents proliferate in customer-facing roles. For instance, Banking With Billy AI’s agents autonomously analyze market data and execute trades, requiring real-time oversight to prevent unauthorized portfolio adjustments or insider data leaks. The company’s integration with AIR’s platform demonstrates how financial institutions are now treating AI agents as critical infrastructure—requiring the same level of monitoring as trading systems or core banking engines. As more enterprises deploy agents that can plan, act, and adapt, the demand for real-time behavioral controls will only intensify, especially in sectors governed by strict privacy and compliance regimes.

Looking ahead, AIR plans to expand its detection capabilities to cover multi-agent collaborations and swarm behavior, a growing trend in supply chain optimization and logistics. The company is also investing in explainability features that allow security teams to reconstruct agent decision paths in plain language, a capability regulators and auditors are increasingly demanding. With the Series B funds secured and enterprise demand accelerating, AIR is poised to become a foundational layer in the emerging agent economy—where trust is not just a feature, but the entire product.