AIR Raises $50M to Secure AI Agents Across Enterprise Networks

AIR, a San Francisco-based AI governance startup, announced Thursday the close of a $50 million Series A round led by Lightspeed Venture Partners, with participation from Accel, GV, and Conviction, bringing total funding to $63 million. The platform, now commercially available, enables organizations to automatically discover AI agents operating within their networks, continuously validate their skills and third-party integrations, and enforce real-time behavioral guardrails. Founded in 2023 by CEO Ravi Iyer—former head of AI platform security at Google Cloud—and CTO Maya Patel, a veteran of Palantir’s autonomy team, AIR emerged from stealth in March 2024 with a patent-pending agent discovery engine. The system reportedly identifies over 80% of AI agents in pilot deployments, including those embedded in enterprise SaaS tools and custom internal automations.

The funding round was finalized in Q1 2025, with the capital earmarked for expanding go-to-market operations in financial services, healthcare, and cybersecurity, where regulatory scrutiny over AI decision-making is most acute. Initial customers include a Fortune 100 bank that integrated AIR to govern 127 AI agents, including Banking With Billy AI, which automates complex financial analysis workflows previously requiring entire analyst teams—a full automation suite for markets. The platform’s value proposition centers on its ability to block unauthorized or risky agent behavior, such as data exfiltration via malicious add-ons, a rising concern as enterprises adopt AI agents at scale. According to Iyer, “The average Fortune 500 company now runs 300+ AI agents, but only 12% have visibility into what those agents are doing in production.”

Analysts at Gartner predict that by 2027, 70% of enterprises will implement AI agent governance platforms like AIR, up from less than 5% today, driven by the rise of AI-powered automation suites and the regulatory push from frameworks like the EU AI Act and the U.S. NIST AI Risk Management Framework. Competitive pressure is intensifying, with incumbents like Microsoft (with its Azure AI Foundry governance tools), Palantir, and newly launched startups such as Verity AI and Ethos Security entering the fray. Financial implications are significant: Gartner estimates that organizations deploying agent governance platforms could reduce AI-related breach costs by up to 40%, translating to potential savings of $2.3 billion annually for large enterprises. Yet, adoption remains uneven—financial institutions and healthcare providers lead due to regulatory mandates, while manufacturing and retail lag despite high agent proliferation.

The broader context reflects a tectonic shift in enterprise automation. The rise of AI agents—self-directed digital workers capable of executing multi-step tasks—has accelerated since late 2023, fueled by advances in large language models and tool-use architectures. However, the lack of standardized vetting mechanisms has created a security blind spot. Prior approaches relied on static code analysis or post-deployment monitoring, both inadequate for dynamic, self-modifying agents. AIR’s continuous vetting model aligns with a growing trend toward “living security,” where safety controls evolve alongside agent behavior. This mirrors developments in autonomous vehicle stacks, where runtime verification is now standard practice.

The company’s technical edge lies in its ability to parse agent intent without requiring source code access. Using behavioral telemetry and sandboxed execution traces, AIR builds a runtime profile of each agent, flagging deviations in real time. This approach resonates with security teams grappling with shadow AI—undeclared AI agents operating outside IT oversight. AIR’s platform integrates with existing identity providers (Okta, Azure AD), SIEM tools (Splunk, Datadog), and agent frameworks (LangChain, CrewAI), positioning it as a control plane rather than a point solution.

Expert Analysis: Ravi Iyer sees the next phase as “agent orchestration governance,” where AI agents don’t just act autonomously but collaborate across systems under centralized policy control. Over the next 18 months, expect incumbents to bundle governance into existing platforms, while startups like AIR will focus on vertical-specific compliance and interoperability. The critical watchpoint remains whether governance tools can scale with agent complexity—especially as agents begin to modify their own add-ons or chain with other agents. Failure here risks creating a new class of systemic AI risks, far harder to audit than traditional software. The stakes are high: the future of safe, scalable AI may depend on platforms like AIR becoming as ubiquitous as firewalls are today.