Vanta agrees to buy Pillar Security to connect AI policies with live protection

Vanta wants the rules companies write for AI agents to shape what those agents can actually do—and produce evidence that the rules hold. On November 9, 2026, it signed a definitive agreement to acquire Pillar Security, whose technology tests AI systems and, the companies say, blocks unauthorized actions while they run.

The companies did not disclose financial terms. Refresh Miami, citing Calcalist, reported an approximate price of $150 million in cash and stock. The announcement concerns an acquisition agreement rather than a completed purchase.

Founded in 2023 by Dor Sarig and Ziv Karliner, Pillar had raised just over $9 million before the agreement, according to Refresh Miami. The startup brings tools for finding, testing and protecting AI applications and agents across their lifecycle.

Connecting written rules to agent actions

Vanta’s platform connects security policies, controls, responsible owners and business commitments. Pillar’s announcement puts Vanta’s reach at more than 20,000 companies. The proposed combination would use that business context to guide security testing and protection, rather than treat an agent’s behavior separately from the organization’s rules.

Vanta describes two capabilities it is acquiring: runtime enforcement, meaning blocking unauthorized actions while an agent operates; and agent red teaming, which uses realistic attacks to find weaknesses before an incident. These address different stages of protection—testing what an agent might do, then enforcing boundaries as it works.

Pillar illustrates the connection with an agent that can access customer records. Before choosing tests or setting operating limits, a security team needs to understand why that agent exists, what it is permitted to do and who owns the risk. The company says those answers shape both its testing and the boundaries it enforces during operation.

The companies intend to connect Pillar with Vanta’s Trust Graph, the platform’s connected business context. Pillar describes three ways information would move through the combined system:

Pillar also brings SAIL, a framework intended to give security, engineering and governance teams a shared way to decide what to protect and where controls belong. The company says SAIL maps 91 AI risks across seven lifecycle phases and has passed 50,000 downloads.

January integration starts with finding the AI

Vanta plans to begin integrating Pillar’s technology in January 2027, according to Refresh Miami. The first tools would identify AI systems across organizations and detect suspicious activity. The outlet also reports that Pillar’s more than 20 employees are expected to join Vanta and form its Israeli research and development center.

Sarig and Karliner are expected to remain and lead AI security development. Sarig has previously described the discovery problem to Refresh Miami through one Fortune 300 customer: after scanning more than 30,000 code repositories, Pillar identified 8,000 AI models and 400 AI systems the customer’s security team had not known existed. Those figures are his account of that customer’s environment.