AI Security Startup AIR Emerges From Stealth With $50M to Secure the AI Agent Supply Chain AI安全初创公司AIR以5000万美元融资秘密推出，致力于保障AI代理供应链安全

TL;DR

• AIR raised $50 million across two seed rounds ($10M from Sequoia, $40M from Greenoaks) to build a security platform for AI agent tooling ecosystems
• The platform discovers active AI agents across enterprise environments, flags unapproved tool usage, and enforces real-time security checks against a continuously updated whitelist
• Approximately 27% of evaluated AI add-ons are currently filtered out by AIR's platform, indicating a significant risk surface in the agent tooling landscape
• Founded by Unit 8200 veterans Yair Saban (CEO) and Niv Hoffman (CTO), with 20+ customers concentrated in regulated sectors like financial services and pharmaceuticals
• AIR positions AI agents as increasingly resembling operating systems, arguing that their tools and plugins lack the verification rigor historically applied to software like device drivers

Why It Matters

As AI agents become embedded in enterprise workflows, the expanding ecosystem of skills, plugins, and MCP servers they rely on represents an unregulated attack surface that traditional security tools are not designed to monitor. AIR's approach addresses a critical gap: treating agent tooling with the same verification discipline that operating systems apply to device drivers and software dependencies.

Technical Details

• Agent Discovery & Monitoring: AIR's platform actively discovers AI agents operating across a company's environment, mapping their tool and plugin usage in real time
• Whitelist-Based Filtering: The system maintains a continuously updated whitelist of approved tools, currently blocking approximately 27% of evaluated add-ons deemed unverified or risky
• Real-Time Enforcement: Security checks are applied at the point of agent action, screening tools before they interact with company systems or the internet
• Target Sectors: Strongest adoption is in regulated industries—financial services and pharmaceuticals—where compliance and audit requirements drive demand

Industry Insight

• The agent tooling security market is rapidly attracting venture capital, with multiple similarly funded competitors (Zenity, Noma Security, Astrix Security, Operant AI) signaling that this is a foundational infrastructure problem, not a niche concern
• Regulated industries will likely serve as the beachhead market, but as AI agent adoption spreads across all sectors, the demand for agent-level security verification will become universal
• The 27% rejection rate of evaluated add-ons underscores that the current AI tooling ecosystem is immature from a security perspective, creating both a risk and a commercial opportunity for platforms that can provide trust and verification at scale

TL;DR

• AIR从隐身状态走出，完成5000万美元种子轮融资（Sequoia领投1000万，Greenoaks领投4000万）
• 平台专注于AI agent生态安全，监控skills、plugins和MCP服务器等工具的使用
• 创始人Yair Saban和Niv Hoffman均来自以色列Unit 8200情报部队
• 平台实时过滤约27%的AI附加组件，已有20+客户，主要来自金融和制药等受监管行业
• 公司约40名员工，计划在美国和欧洲扩展研究与商业化

为什么值得看

AI agent正迅速成为企业核心基础设施，但其工具链安全缺乏有效监管机制。AIR的解决方案填补了这一关键空白，为受监管行业提供了必要的合规保障。

技术解析

AIR平台能够发现企业环境中的活跃agent，标记未授权工具使用，并实时执行安全检查。通过持续更新的白名单机制，目前过滤掉约27%的评估附加组件。平台将AI agent安全视为持续的基础设施问题，而非简单的扫描任务。

行业启示

AI agent安全正成为独立赛道，预计将涌现更多类似Zenity、Noma Security的竞争对手。受监管行业（金融、制药）将成为AI安全产品的首批采用者，验证商业模式后向其他行业扩散。传统软件验证理念需要延伸至AI agent生态，建立新的安全标准。