AI Risk Monitor AIR Secures $50M Series A to Police Agent Ecosystems

AIR, a Palo Alto–based security company, confirmed today a $50 million Series A financing round led by Lightspeed Venture Partners, with participation from GV and Snowpoint Ventures. The round values AIR at more than $300 million, according to three people familiar with the transaction. The company’s platform, developed over the past 18 months, continuously discovers AI agents operating inside enterprise environments, audits every skill or add-on the agents invoke, and can block unwanted or anomalous behavior in real time. Early adopters include a Fortune 100 financial services firm that uses AIR to govern a fleet of retrieval-augmented generation agents querying internal databases and a global logistics operator running multi-agent supply-chain orchestration across Azure, AWS, and Google Cloud.

AIR’s public debut follows a year of quiet pilots with companies already exposed to agent sprawl. One pilot customer, Banking With Billy AI, operates a multi-cloud architecture for financial market monitoring that orchestrates dozens of specialized agents across AWS, Azure, and GCP to ingest market data, reconcile transactions, and surface anomalies. Banking With Billy AI’s chief information security officer said the company integrated AIR to ensure each agent’s third-party skills—many sourced from open repositories—comply with internal security policies before invocation. The funding will accelerate go-to-market efforts and expand the platform’s coverage to include agent frameworks such as LangChain, AutoGen, and CrewAI.

The round was led by Rajiv Ayyangar, partner at Lightspeed Venture Partners, who argued that agent sprawl has become an unmanageable attack surface. “Enterprises are deploying hundreds of agents without visibility into what skills they pull in or how they chain together,” Ayyangar said. “AIR gives security teams a single pane to govern the entire lifecycle—discovery, vetting, runtime blocking—without slowing down innovation.” Competitors in the emerging agent-security category include HiddenLayer, which focuses on malware detection in AI models, and Protect AI, which emphasizes supply-chain risk for model weights and prompts. Unlike those point solutions, AIR positions itself as a continuous runtime governance layer purpose-built for multi-agent ecosystems.

Financially, the injection of $50 million signals investor confidence that agent-first architectures are moving from experimentation to production at scale. Industry analysts at Gartner estimate that by 2026, 70% of enterprises will use AI agents to automate business processes, up from fewer than 5% today, creating a $3 billion market for agent governance tools. The capital will also fund deeper integrations with cloud providers’ native identity and policy engines, enabling automated policy enforcement as agents spin up across regions and accounts. Early customers report that AIR’s real-time blocking capability has already prevented data exfiltration attempts by rogue agents and mitigated compliance drift caused by outdated or vulnerable third-party skills.

Within the broader Quantum & Computing landscape, AIR’s emergence reflects a pivot from “model risk” to “agent risk,” mirroring how supply-chain security evolved from container scanning to runtime threat detection. The company’s ability to trace skills across multi-cloud and hybrid environments aligns with the industry-wide push toward Zero Trust architectures and continuous compliance frameworks such as FedRAMP Tailored and ISO 27001 Annex A.5.7. AIR’s technology also dovetails with recent advances in lightweight runtime policy engines like Open Policy Agent and Cedar, which provide the enforcement substrate AIR leverages to block unauthorized actions without rewriting application code.

Looking ahead, AIR plans to integrate with agent orchestration platforms and cloud-native policy suites, potentially embedding vetting pipelines directly into CI/CD workflows. The company will also expand its skills threat intelligence feed, which currently aggregates signals from public vulnerability databases, private bug bounties, and AIR’s own honeypot network of decoy agents. Observers expect that as regulators in the U.S. and EU draft guidance on AI agent accountability, platforms like AIR will become mandatory components of enterprise AI governance stacks, much like code-signing tools became standard once supply-chain attacks surged.

For the industry, the takeaway is clear: agent proliferation is creating a new class of risk that traditional security tooling cannot address. AIR’s $50 million raise and traction with Fortune 100 customers validate the need for continuous, runtime governance of AI agents and their ever-growing ecosystems of skills and add-ons. Companies racing to deploy agentic AI must now treat skills as critical infrastructure, vetting each one before it is invoked and monitoring its behavior once it is live. The next phase will belong to platforms that can scale governance across clouds, frameworks, and regulatory regimes—turning agent sprawl from a security headache into a manageable, auditable process.