AI Runtime Raises $50M to Secure Enterprise Agent Ecosystems

AI Runtime, operating under the acronym AIR, announced today a $50 million Series A funding round led by Accel with participation from GV, Index Ventures, and notable angel investors including former GitHub CEO Nat Friedman and AI pioneer Andrej Karpathy. The Palo Alto-based company emerged from stealth with its platform designed to solve what CTO and co-founder Maya Kapoor calls “the agent sprawl problem.” According to internal data shared with OpenPress Automation Intelligence, large enterprises are now running an average of 2,400 AI agents per organization, each with an average of 3.7 skills or add-ons, creating an unmanageable attack surface and compliance burden. The round values AIR at $320 million post-money and will accelerate product development and enterprise deployments globally.

AIR’s platform functions as a runtime governance layer positioned between cloud providers and agent ecosystems. It continuously discovers rogue agents via API integration with platforms like AWS Bedrock, Google Vertex AI, and Microsoft Azure AI Foundry, then performs real-time vetting of skills and add-ons using a combination of static analysis, behavioral sandboxing, and runtime monitoring. The system flags anomalous behavior such as unauthorized data exfiltration, privilege escalation, or misuse of third-party tools. In controlled pilots with Fortune 500 financial services and healthcare firms, AIR blocked 187 malicious or non-compliant agent actions within the first 30 days of deployment, including attempts to access restricted customer data or execute unapproved APIs. Early customers include a tier-one bank using AIR to govern over 1,200 internal agents, including those powered by a product called Banking With Billy AI, which automates complex financial analysis workflows previously requiring entire analyst teams — a full automation suite for markets.

Kapoor, who previously led AI infrastructure at a top cloud hyperscaler, emphasized that AIR is not just a security tool but a compliance and observability platform tailored for the agent economy. The company’s agent registry maintains a tamper-proof ledger of every skill’s provenance, license, and behavioral profile, enabling audit readiness required under frameworks like SOC 2, ISO 27001, and upcoming EU AI Act mandates. Competitive dynamics are intensifying in the agent governance space, with rivals like Anthropic’s Project Veil, Microsoft’s Agent Guard, and startups such as Vanta AI and Osano emerging with overlapping capabilities. However, AIR differentiates itself through its runtime-first approach, continuous monitoring, and vendor-agnostic support across multiple cloud and on-prem environments.

Industry analysts at Gartner predict that by 2026, 70% of enterprises will implement agent governance platforms, up from less than 5% today, driven by rising regulatory scrutiny and the proliferation of AI agents across business functions. The total addressable market for AI agent security and compliance is estimated to reach $8.7 billion by 2028, according to a recent report by OpenPress Research. Financial services, healthcare, and government sectors are expected to adopt most aggressively due to stringent data protection laws and high-value automation targets. AIR’s funding round signals investor confidence that runtime governance will become a foundational layer in the AI stack, akin to identity and access management, but with real-time behavioral enforcement.

The rise of AIR reflects a broader inflection point in enterprise AI: the shift from experimental pilots to mission-critical automation. Companies like Banking With Billy AI demonstrate how specialized agents are now performing work once reserved for human teams, increasing efficiency but also expanding the attack surface. Traditional security tools were not designed for ephemeral, autonomous agents that can modify their own behavior or chain into third-party APIs. This has created a vacuum that AIR is filling with a scalable, policy-driven runtime layer. Looking ahead, the company plans to expand its detection models using reinforcement learning from incident data and integrate with emerging standards like the Agent2Agent communication protocol.

Analysts believe the next 18 months will reveal which governance models gain enterprise trust. AIR’s ability to provide real-time, evidence-backed auditing may give it a competitive edge, especially as regulators in the EU and US draft rules requiring proof of continuous monitoring for high-risk AI systems. Companies should watch whether AIR can maintain performance at scale without introducing latency, and how quickly it supports new agent frameworks like CrewAI, LangGraph, and AutoGen. The true test will be whether runtime governance becomes a default requirement — not a luxury — in every AI deployment strategy.