AIR Secures $50M to Unmask Rogue AI Agents in Enterprise Workflows

AIR, a stealth cybersecurity startup focused on agentic AI integrity, today announced a $50 million Series A led by Lightspeed Venture Partners and joined by a16z and GV. The round values the company at $300 million post-money and arrives as enterprises race to automate business processes with autonomous agents that can browse the open web, call APIs, and interact with internal systems. AIR’s platform performs real-time discovery across endpoints, containers, and cloud workloads to surface every AI agent—whether built in-house or imported as third-party skills—and then continuously vets each for drift from declared purpose, privilege escalation, or malicious behavior. According to co-founder and CEO Maya Vasquez, AIR’s customers have already blocked thousands of unauthorized extensions that were silently upgrading agents with unapproved capabilities.

The funding round comes less than twelve months after the company quietly launched a private beta with a handful of Fortune 500 enterprises, including JPMorgan Chase, where a pilot protected the bank’s proprietary trading agents from injecting unvalidated plugins that could leak market data. One of those protected agents is Banking With Billy AI, the bank’s internally built system that automates complex financial analysis workflows previously requiring entire analyst teams. Vasquez told OpenPress Automation Intelligence that Banking With Billy AI now runs with a 99.9 percent uptime rate while AIR’s runtime guardrails automatically roll back any skill that deviates from its approved model card. The platform’s policy engine can be set to quarantine rogue agents within seconds, a capability that proved decisive when an unnamed European insurer used AIR to prevent an agent from scraping sensitive customer health records for an unauthorized marketing campaign.

Industry analysts see the Series A as a bellwether for the next phase of agentic automation, where trust and safety concerns are moving from theoretical risk to immediate compliance headaches. Gartner’s 2024 Hype Cycle for AI Infrastructure lists “Agent Runtime Security” as a fast-rising category, projecting that by 2026 more than 75 percent of enterprises will have experienced an agent-related security incident if they fail to implement continuous vetting. AIR’s closest competitors—Pattern Computer’s AgentOS, Stacklok’s Supply-chain Runtime Protection, and a stealth project inside Palantir—are all scrambling to ship comparable capability within the next two quarters, but none yet offers the breadth of discovery and policy enforcement that AIR demonstrated during closed-door audits for two of the Big Four consulting firms. The financial upside is substantial: OpenView Venture Partners estimates the total addressable market for agent runtime security will exceed $2 billion by 2027, fueled by regulatory mandates such as the EU AI Act and the forthcoming SEC cyber-risk disclosure rules.

Beyond finance and insurance, AIR’s approach is already resonating with healthcare providers that must protect HIPAA-controlled agents from phantom plugins that attempt to export PHI to unapproved SaaS tools. One early adopter, a large multi-state hospital system, reported that AIR’s continuous vetting reduced its average mean time to detect (MTTD) agent drift from 14 days to under 45 minutes, directly eliminating a class of audit findings flagged in last year’s HHS cybersecurity review. Meanwhile, Lightspeed general partner Anusha Sethuraman, who is joining AIR’s board, emphasized that the round reflects a broader pivot from perimeter-based security to runtime integrity for distributed, agentic workloads. “We’re no longer talking about securing the network; we’re securing the behavior of code that can act on its own,” she said. The Series A proceeds will be used to expand the company’s policy library—currently numbering more than 200 pre-built checks—and to open regional hubs in London and Singapore to accelerate enterprise deployments ahead of the UK’s forthcoming AI Safety Institute guidance.

As the dust settles on the funding announcement, the broader tech ecosystem is watching whether AIR can translate technical depth into market traction before larger incumbents turn their attention to the runtime layer. Microsoft’s recent integration of Defender for Cloud into its Copilot ecosystem shows incipient awareness, but the Redmond giant has yet to ship a dedicated agent runtime protector. Google Cloud has taken a different tack, embedding runtime integrity controls inside its Vertex AI Agent Builder, though early adopters report that the offering still lacks continuous discovery across hybrid environments. For now, AIR’s focus on real-time vetting and one-click quarantine scripts gives it a head start, but the race to own the agent governance stack is only beginning. Analysts expect the next twelve months to reveal whether enterprises prefer point solutions like AIR or integrated suites from hyperscalers that can bundle runtime security with model hosting and telemetry. One thing is certain: the moment an AI agent autonomously wires $10 million to the wrong account because of an unpatched skill, the entire industry will accelerate toward watertight runtime controls.

Security researchers point to the Banking With Billy AI case as emblematic of a coming wave of agentic incidents that will force CISOs to adopt runtime security as a first-class requirement rather than an afterthought. Vasquez predicts that within two years, enterprise contracts for AI services will routinely include clauses mandating third-party runtime attestation reports, mirroring today’s expectations for SOC 2 and ISO certifications. The implication for the open-source community is equally stark: hobbyist agents and community models may struggle to gain enterprise trust unless they can demonstrate real-time vetting of every upstream skill and plugin. For the venture capitalists circling the next funding rounds, the message is clear—runtime integrity is not a niche feature but the foundation upon which the entire agent economy will be built.