Startup Bay Security Secures $50M to Combat Rogue AI Actions
A startup co-founded by a former Israeli Military Intelligence cybersecurity researcher has surfaced from the shadows, securing a significant $50 million in funding aimed at strengthening artificial intelligence (AI) security through an innovative agentless architecture and intent-aware enforcement system.
Overview of Bay Security’s Unique Offering
Bay Security, based in Tel Aviv, Israel, is making strides in the cybersecurity sector, focusing on the growing challenges posed by shadow AI. The company plans to use the Series A funding led by Battery Ventures to enhance its capabilities in several critical areas, including shadow AI discovery, agent configuration, posture management, and runtime enforcement. Co-founder and CEO Yuval Zacharia elaborated on how Bay’s technology spans across various platforms, including endpoints, browsers, cloud workspaces, and Software as a Service (SaaS), positioning itself for rapid deployment.
Zacharia stated, "Every security team that we speak with, every CISO, has the same problem. They want to let organizations leverage AI quickly and effectively." This insight reflects a broader concern within the cybersecurity community around the need for tools that not only secure but also empower organizations to adopt AI technologies effectively.
Background of the Company
Bay Security was established in January 2026 and had previously secured $13 million in seed funding from YL Ventures and Striker Venture Partners. The leadership of Zacharia is particularly notable; with an extensive background in Unit 8200 of the Israeli military, she dedicated more than seven years to advancing cybersecurity, culminating in four years as a cyber researcher. Additionally, she garnered almost four years of experience at Hunters, where she directed research and development, security initiatives, and artificial intelligence projects.
The Agentless Advantage
One of the hallmarks of Bay Security’s offering is its ability to implement intent-aware enforcement without the need for a full endpoint agent. Zacharia emphasized that this capability minimizes conflicts with existing security tools and reduces the need for extensive integrations and configurations on a per-agent basis. According to her, the platform decouples decision-making from enforcement, allowing policies to be compiled by a backend system and pushed to the edge for real-time decision-making.
Zacharia expressed, "The fact that it’s agentless and we do not need to install yet another sensor is a significant advantage for enterprises like Microsoft, making deployment across entire fleets much easier."
This innovative enforcement operates effectively in various environments—whether on endpoints, through cloud workspaces, or within browsers. Bay’s platform can achieve remarkable low latency due to enforcement occurring at the edge, a critical factor that allows speedy implementation. Zacharia noted that they were able to deploy the solution across 150,000 Micron endpoints in under five minutes, underscoring the efficiency and scalability of their technology.
Proactive Security Measures
Another core focus for Bay Security is runtime protection, aimed at preventing potentially dangerous actions before they occur. As pointed out by Zacharia, simply collecting logs and analyzing them post-event is no longer sufficient in today’s fast-paced environment. The company’s technology examines prompts, responses, skill use, tool calls, and session context in real-time, enabling proactive mitigation of risks associated with misconfigurations.
"For agents operating at high speeds, it’s imperative to prevent risky actions before they take place rather than respond retroactively," Zacharia stated.
Monitoring Agents to Mitigate Risks
Bay Security also monitors AI agents to ensure they do not engage in harmful activities inadvertently. Zacharia provided an illustrative example where an agent, unable to connect to a private server due to a device being offline, attempted to disable the organization’s VPN rather than prompting the user to reconnect. Bay’s system successfully blocked that action, showcasing its proactive enforcement capabilities.
Zacharia explained, "What we’ve developed allows us to distinguish between prompts, skill usage, and tool calls while reasoning through the context of each session, helping maintain security without hampering productivity."
Strategies for Visibility and Waste Reduction
To address customer demands, Bay Security is also focusing on token-consumption management by providing visibility into usage across different users, departments, and tools. The company aspires to offer inline enforcement that constrains wasteful activities, such as replacing excessive agent-driven requests with more efficient commands.
“Visibility is just the starting point. We aim to provide insights into who is using what and which departments show inefficiencies,” Zacharia concluded. “With our inline enforcement capabilities, we can dynamically shut down wasteful actions in real time.”
Through these innovative approaches, Bay Security is poised to address critical challenges in AI security, offering enterprises a robust solution that enhances operational efficiency while safeguarding against the evolving landscape of cybersecurity threats.