Bay Emerges From Stealth With $50M to Guard Against Rogue AI Agent Actions

Israeli cybersecurity startup Bay launched publicly on Tuesday with $50 million in total funding. The capital consists of a $37 million Series A round led by Battery Ventures and a previously undisclosed $13 million seed round led by YL Ventures and Striker Venture Partners. Micron Ventures and Wiz Chief Executive Assaf Rappaport also invested.

Bay was founded nine months ago by Chief Executive Yuval Zacharia, Chief Product Officer Aner Morag, and Vice President of R&D Boris Danilovich, all veterans of the Israel Defense Forces' Unit 8200. Danilovich previously spent six years at CrowdStrike, while Zacharia led cyber intelligence operations before working in industry AI security research.

Unlike first-generation AI safeguards that primarily filter prompts and chat outputs, Bay focuses on operational actions. Its scanner maps all local models, extensions, and agents running on an employee device. Whenever an agent attempts to execute an action, a local binary called Bay Enforcer evaluates the request against automated policies generated from past activity, delivering a verdict in milliseconds.

The startup already counts Micron Technology, Wix, ClickUp, and Peloton among its enterprise customers. Bay claims that Fortune 100 clients tripled their active use of autonomous agents within 30 days of deploying the platform. The new funds will support sales expansion in the United States and Europe alongside engineering growth in Tel Aviv.

As software development and workplace productivity shift from static chatbots to autonomous agents that hold system privileges, the risk profile changes drastically. An agent given broad API keys can follow an innocent instruction and accidentally wipe databases or open servers to the public internet.

By evaluating agent behavior on the host machine before any command runs, security teams can avoid bottlenecking employee rollouts. This runtime control model aims to treat agent governance much like firewalls treated enterprise networks during the early days of corporate internet adoption.