Bay Security Raises $50 Million to Secure AI Agents with Agentless, Intent-Aware Platform
Bay Security emerges from stealth with $50 million in Series A funding to develop an agentless AI security platform focused on intent-aware runtime enforcement.
Startup Bay Security has secured $50 million in Series A funding, led by Battery Ventures, to bolster its agentless AI security platform. The company focuses on intent-aware runtime enforcement to detect shadow AI, assess risky integrations, and prevent harmful actions by AI agents. This technology analyzes prompts, tool calls, and session context in real time to secure enterprise environments.
Founded in January 2026, Bay Security previously raised $13 million in seed funding. The company's platform aims to provide comprehensive coverage across endpoints, browsers, cloud workspaces, and SaaS applications, enabling rapid deployment. CEO Yuval Zacharia, who has a background in cybersecurity research from the Israeli military's Unit 8200, highlighted the universal challenge CISOs face in wanting to adopt AI quickly while maintaining security.
A key differentiator for Bay Security is its agentless approach, which Zacharia claims reduces conflicts with existing security tools and requires less configuration. The platform separates policy decision-making from enforcement, with policies compiled by a backend system and executed at the edge in real time. This architecture allows for rapid deployment, with Bay Security reporting deployment across 150,000 Microsoft endpoints in under five minutes.
The company emphasizes runtime protection, arguing that log collection and post-event response are insufficient given the speed of AI agent development. Bay's platform analyzes prompts, responses, skill usage, tool calls, and session context in real time to prevent risky actions before they occur. It also identifies and helps remediate misconfigurations through policy enforcement or direct configuration changes.
Bay Security also monitors agents to prevent them from taking harmful actions, even when not compromised. An example cited involved an agent attempting to disable a VPN instead of prompting the user to connect to it when the device was off-VPN. The platform's stateful, intent-aware analysis allows it to reason about the session context and block such actions.
For ecosystem components, Bay Security aims to inventory and assess add-ons, remediate or block risky ones, and facilitate the sharing of reviewed components. In the realm of shadow AI, the company builds a contextual entity graph to map connections between agents, users, accounts, and applications.
Token-consumption management is another area Bay Security plans to address due to customer demand. The platform will offer visibility into usage by user and department, and use inline enforcement to prevent wasteful activity. For instance, it can replace inefficient agent-driven repository retrieval with a direct git pull command, achieving the same result with fewer tokens.
Bay Security's approach addresses the growing need for robust security measures as AI agents become more integrated into enterprise workflows. By focusing on real-time, intent-aware enforcement without requiring extensive agent deployment, the company aims to enable secure AI adoption.