Tel Aviv AI Security Startup Bay Raises $50 Million and Expands Hiring

Tel Aviv-based AI security company Bay emerged from stealth on November 10 with $50 million in total funding, including a new $37 million Series A and a previously undisclosed $13 million seed round. It plans to expand product and engineering, and its careers page lists development and research roles in Tel Aviv.

The company is building controls for businesses that allow AI agents to carry out tasks inside workplace systems. Its proposition moves beyond deciding who can access a system to examining what an agent does with that access. Bay says it can allow an action, stop it or refer it for human approval under the organization's policies.

There are two local angles: a company based in the city is expanding, and employers adopting AI have another example of the security products being developed around it. The financing is documented in Bay's announcement. Claims about the platform's capabilities remain the company's account, rather than the findings of an independent product test conducted for this article.

The funding total covers two separate rounds

Bay says Battery Ventures led the Series A, with participation from Micron Ventures, Microsoft's M12 venture fund and Wiz CEO Assaf Rappaport. YL Ventures and Striker led the seed round. The distinction between the two rounds matters: the $50 million headline is the combined funding total, not the size of the new Series A alone.

The company intends to expand its product and engineering teams and grow sales, marketing and customer operations in the United States and Europe. Globes reports that Bay was founded in early 2026. For Tel Aviv readers, the concrete employment question is which activities and roles are being developed locally alongside those overseas operations.

The announcement does not provide a valuation, a profitability figure or a promised number of new Tel Aviv hires. Funding and revenue are different measures, too. Capital can support development and expansion without establishing how much a company earns from customers or how long those customers will continue using its product.

Why an agent's permissions do not explain its whole task

An AI assistant that drafts an email can leave the act of sending it to a person. An agent equipped with tools and permissions may perform parts of a task itself. As a hypothetical example, a team might let an agent edit files or run a test. The operational question is which actions are appropriate on the way to the assigned goal.

Bay places its product at that point. Its enforcement page describes examining the wider work session to interpret an action in context, then applying company policy as the agent acts. That is the approach Bay offers; the description does not prove that it will recognize every unsafe situation in every environment.

The distinction is useful even before a business chooses a supplier. Permission to edit a file is not necessarily an instruction to alter every file a tool can reach. Teams need to define the permitted task, the actions requiring approval and the point at which work should stop. Bay seeks to automate part of that oversight while work is happening.

Recording the sequence and testing rules before enforcement

On its investigation product page, Bay says it records the agent's work session and distinguishes human actions from autonomous ones. Its stated purpose is to help security teams reconstruct how a task progressed, rather than seeing only the final action. That puts an emphasis on explaining what was done and on whose behalf.

The company also describes a simulation option for testing policies before applying them to users. In practical terms, a business could ask to see which actions a proposed rule would allow or block. This is an advertised capability, not independent evidence of how many mistaken blocks the system prevents.

A prospective customer could prepare a familiar task containing an allowed action, one requiring approval and one that should stop. Useful questions then follow: what appears in the record, who receives the approval request and how does work resume? These are suggested evaluation steps, not claims that a particular customer has already performed such a test.

What the Tel Aviv vacancies actually show

Bay's careers page, checked on November 11, lists Tel Aviv positions for Backend Developer, Frontend Developer, Full Stack Developer and DevOps Engineer, together with Head of Research and Security Researcher. The same page lists sales roles in the United States. This is a snapshot of advertised positions and can change as recruitment progresses.

For local candidates, the relevant job description is more useful than the funding headline. Check the required experience, responsibilities, location and working arrangement before applying. A Tel Aviv location label does not itself confirm hybrid work, remote eligibility or salary. One listed title also does not establish how many people the company intends to hire into it.

The founders named on Bay's website are CEO Yuval Zacharia, Chief Product Officer Aner Morag and Vice President of Research and Development Boris Danilovich. Candidates can use the published leadership information to understand the organization, then assess a role against its actual requirements. The announcement offers no guarantee of employment or individual career outcome.

What businesses and applicants can check next

A business considering this kind of system can begin with an inventory of the AI tools its staff use and the tasks those tools are allowed to perform. Questions about coverage, record storage, access to records, approvals and effects on everyday work can then be tied to a real workflow. The public material reviewed does not quote a price suitable for comparing costs across organizations.

Bay names Micron, Peloton, ClickUp and Wix among its customers. Those are company-reported customer relationships, not a substitute for evaluating another workplace. Two organizations using the same AI tools may give them different permissions and assignments, so a familiar customer name does not establish the result a new customer should expect.

For applicants, the next step is to follow the live vacancies. For employers, it is to define a test case and examine how the proposed controls handle it. Those steps turn a financing announcement into questions that can be answered: which work is expanding in Tel Aviv, and what control remains with a business as it delegates more tasks to software?