Bay Security gets $50 million to stop dangerous AI agent steps

Bay Security, a startup in Tel Aviv, Israel, has $50 million in Series A funding, and Battery Ventures is the lead investor. The software checks the work of AI agents and stops dangerous tasks while they occur. The chief executive set up the company in January 2026, and it got $13 million in seed funding before.

How it works

Bay checks prompts, tool use and session context while an agent works, and stops a dangerous task before it starts. It puts no agent software on each device. A backend system makes the policies and sends them to the edge, where decisions occur.

The proof

The chief executive tells that the software was put on 150,000 Microsoft endpoints (devices) in less than five minutes. In one example, an agent tried to stop the company VPN (a safe connection to a company network), and the software stopped the task. These are claims of the company.

Open questions

The source gives no tests by external persons and no number of customers. Bay also has a plan to stop AI use that is not necessary, but this is only a plan.

Insight: Agent control is becoming its own market

Bay has $63 million in total funding: $13 million in seed funding and $50 million in Series A funding. CrowdStrike makes agents that look into cybersecurity incidents, and Microsoft has a plan for agents that monitor agents. This reads as a race to be between each agent and its tools, because the one that holds that point sees all agent tasks. Investors possibly think that this is a new market.