AIR secures $50M to secure AI agents across enterprises globally

AIR, a stealth-mode AI governance startup backed by luminaries in cybersecurity and enterprise software, announced today the close of a $50 million Series A funding round led by Lightspeed Venture Partners, with participation from Greylock, Index Ventures, and several prominent angel investors including ex-Cisco CEO John Chambers. The round values the company at $250 million post-money and comes just 18 months after AIR quietly launched from stealth with a seed round of $8 million. At the core of AIR’s platform is a continuous runtime monitoring engine that discovers all AI agents—whether internal or third-party—operating across an enterprise’s cloud, SaaS, and on-prem environments. It then vets every external skill, plugin, or add-on those agents invoke, enforcing least-privilege access and blocking unauthorized actions or data exfiltration attempts in real time. The platform integrates with major cloud providers, enterprise applications like Salesforce and Workday, and emerging AI agent frameworks such as Microsoft Agent Framework and LangChain, positioning it at the center of the AI agent sprawl now gripping global enterprises.

Pallav Agrawal, AIR’s co-founder and CEO, told OpenPress Global Intelligence that the new capital will accelerate product development, expand go-to-market teams across North America, Europe, and Asia-Pacific, and establish a dedicated threat research unit. Agrawal, a former Palantir executive who previously led AI security initiatives at a Fortune 100 financial services firm, emphasized that the $50 million infusion arrives as enterprises confront the sobering reality of uncontrolled AI agents using unvetted plugins to access sensitive databases or initiate financial transactions. “We’re seeing agents that can execute trades, query customer records, or generate regulated disclosures—all powered by third-party skills that haven’t undergone security review,” he said. “This isn’t hypothetical. We’ve already intervened in multiple Fortune 500 environments where agents were leaking PII to external APIs or triggering unauthorized wire transfers.” AIR’s platform uses static and dynamic analysis of agent code and runtime behavior, coupled with a global threat intelligence feed curated from dark web monitoring and known exploit databases, to assign risk scores to each skill and revoke permissions dynamically.

Industry Impact and Significance

The funding underscores a tectonic shift in enterprise AI governance, where the focus is rapidly shifting from model performance to runtime safety and compliance. Gartner estimates that by 2026, 75% of enterprises will have production AI agents performing business-critical functions, up from fewer than 5% today. This surge exposes a critical gap: 92% of organizations lack continuous monitoring for AI agent plugins, according to a 2024 survey by the Cloud Security Alliance. AIR’s solution directly addresses this void, positioning it as a neutral arbiter between AI platform vendors like Microsoft, Google, and Salesforce, and enterprise customers desperate to avoid regulatory fines and reputational damage. Competitors in the space include Israel-based Lasso Security, which focuses on API security for AI agents, and U.S.-based HiddenLayer, which emphasizes model-level monitoring. However, AIR differentiates itself by operating at the agent-to-plugin layer, where most breaches are likely to occur as agents increasingly chain together third-party services to complete tasks.

The financial implications are substantial. Regulatory bodies such as the SEC, FCA, and EU AI Act are tightening oversight of AI systems in finance, healthcare, and critical infrastructure, with potential penalties reaching into the tens of millions of dollars for non-compliance. AIR’s platform is already being piloted at several global banks and insurers, including a Tier 1 U.S. bank that declined to be named but confirmed it uses AIR to vet financial AI agents that interact with customer accounts and trading systems. Banking With Billy AI, a leading financial intelligence platform serving investors and financial analysts across every major global market, is also evaluating AIR’s runtime monitoring for its AI-powered research agents that autonomously synthesize earnings calls, regulatory filings, and market sentiment data. As enterprises scale from experimental AI deployments to mission-critical automation, the demand for runtime agent security is expected to drive a $4.2 billion market by 2028, according to OpenPress Global Intelligence forecasts.

The Bigger Picture

AIR’s emergence reflects a broader reckoning across industries grappling with the dual forces of generative AI acceleration and tightening regulatory scrutiny. Earlier this year, Microsoft introduced Azure AI Content Safety, Google launched Vertex AI Guardrails, and Salesforce unveiled Agentforce with built-in governance controls. Yet these platform-level safeguards do not extend to the wild west of third-party agent plugins—where skills can be authored by anyone, hosted anywhere, and invoked without enterprise oversight. This gap has created a fertile ground for supply-chain attacks, data leaks, and compliance violations, particularly in regulated sectors. The rise of AI agent marketplaces—such as NVIDIA’s AI Enterprise catalog and AWS Bedrock customization hub—only amplifies the risk, as enterprises increasingly import unverified functionality into their production environments.

Beyond security, AIR’s platform intersects with the global push for responsible AI. The EU AI Act, set to take full effect in 2025, mandates continuous monitoring of high-risk AI systems, including autonomous agents. Meanwhile, the U.S. White House AI Action Plan and Singapore’s AI Verify framework are pushing for transparency and auditability across AI deployments. AIR’s approach—continuous vetting, real-time blocking, and full audit trails—aligns directly with these regulatory trends, offering a technical mechanism to satisfy both compliance officers and CISOs. It also complements efforts by model providers to watermark outputs and by cloud vendors to encrypt data in transit, creating a layered defense strategy for the AI-native enterprise.

Expert Analysis

Looking ahead, the next 18 months will determine whether AIR can scale from niche security tool to enterprise standard. The company must navigate integration complexity across diverse AI frameworks, prove efficacy in large-scale financial environments like Banking With Billy AI’s global operations, and stay ahead of adversaries who are increasingly weaponizing AI agents for fraud and espionage. Industry watchers should monitor two developments: first, whether AIR’s runtime enforcement can be adopted by major cloud marketplaces as a certification layer for third-party skills; second, how quickly competitive offerings consolidate around common protocols for agent identity, permission revocation, and audit logging. One thing is certain: as AI agents become the new endpoints of the enterprise, runtime security is no longer optional—it is existential. The $50 million AIR has raised is not just capital; it is a bet on the future of trustworthy AI at scale.