AI in Israel

Vanta signs deal to acquire Israel’s Pillar Security, with AI agent controls planned for 2027

Calcalist values the deal at about 150 million dollars. Vanta plans monitoring in January 2027 and runtime enforcement in the summer.

A blue security gateway around abstract agents and document paths
AI-generated conceptual illustration of agent security, not a depiction of an actual product or event.

Vanta announced on November 9, 2026 that it had signed a definitive agreement to acquire Israeli AI security company Pillar Security. Calcalist reported a value of approximately 150 million dollars in cash and shares. The companies confirmed the agreement but did not disclose the consideration in their announcements. This is a signed transaction, rather than an announcement that the acquisition has closed.

The planned combination brings together two tasks that organizations sometimes manage separately: setting rules for artificial intelligence and enforcing them while systems use data and tools. Vanta plans to integrate Pillar's capabilities during 2027. For customers, that distinction matters. The announcement sets out a roadmap for a combined product; it does not mean every capability is available now.

What Vanta is buying, and what remains unfinished

Ziv Karliner and Dor Sarig founded Pillar in 2023. In its own announcement, the company describes technology that discovers AI applications and agents, tests them and restricts unauthorized actions in real time. Vanta contributes its Trust Graph, which connects policies, controls, accountable owners and business risks. The intention is to give security alerts organizational context that teams can act on.

According to Calcalist, Pillar will become Vanta's Israeli research and development center after the transaction closes. More than 20 employees are expected to join, with the founders continuing to lead the AI security business. These details also describe plans for after completion. The announcements do not supply a closing date that establishes exactly when the organizational transition will finish.

Vanta says it serves more than 20,000 customers. That figure explains the potential distribution channel for Pillar's technology, but it does not show how many customers already use it or will buy the combined offering. Access to an established customer base is a commercial opportunity. Actual adoption, deployment scale and benefits for each organization remain to be seen.

Policies need to reach the point where an agent acts

The problem the deal seeks to address begins when an agent receives permission to perform an action. A document can prohibit sending customer data to an outside channel, but preventing that transfer requires a control where execution occurs. Detecting an incident afterward, blocking it beforehand and keeping a record are different functions. A system that performs one does not necessarily perform all three.

Vanta describes attempted access to restricted information or the sharing of sensitive financial data through an unapproved channel as examples of the planned capability. The company says the integration should block an unauthorized action, notify the responsible owner and produce an audit trail. This is a supplier's capability claim, rather than proof that the system prevents every leak in every environment or replaces sound permission settings.

Business context is the interesting part of the combination. The same request to read a customer record might be legitimate for a service agent and unnecessary for one summarizing marketing material. A decision requires knowing the agent's role, who approved it and which data it needs. That points to the operational significance: connecting a technical decision to a defined business rule and an accountable owner.

January monitoring, summer enforcement

Vanta's roadmap schedules inventory capabilities for January 2027, covering applications, models, agents and MCP servers. MCP is a protocol connecting AI tools with data sources and other tools. The inventory is intended to associate these systems with owners, approved uses, data permissions and risks. Runtime monitoring and the detection of suspicious activity or safeguard violations are also planned for that month.

Adversarial testing, commonly called red teaming, tailored to an organization's context is planned for spring 2027. Runtime enforcement is scheduled for summer 2027. That separation is particularly important in purchasing decisions: seeing a violation is different from stopping it. A customer seeking immediate blocking should establish what exists today, what requires a separate product and what remains on the development roadmap.

Vanta says AI Governance will be a paid offering in January 2027, but its acquisition announcement does not provide a price list. Customers therefore cannot assume the capabilities will automatically be included in every existing subscription. Pricing, integration terms, support and a firm availability date belong in a commercial proposal. A published roadmap helps planning but does not guarantee identical deployment timing for every customer.

What organizations can check now

A practical starting point is an inventory of agents already operating and the permissions they hold. Each should have a clear purpose, an accountable owner and an understandable boundary on its actions. This recommendation follows from the proposed design: a relationship map is useful only if it reflects the organization as it operates, rather than an outdated list missing new connections or forgotten service accounts.

An evaluation should include a permitted action that looks unusual, to assess whether controls interrupt legitimate work, as well as an unauthorized attempt in a test environment. Teams should establish who receives alerts, who can release a block, what information is logged and how access can be revoked. These are product evaluation questions, not findings of a problem at either company or among their customers.

Security and compliance leaders share responsibility here. A record of a blocked action can help demonstrate that a control worked in a particular instance. It does not alone prove overall compliance with every standard or legal obligation. The potential value is a clear connection between a rule, an action and evidence. Organizations first need agreement on what is prohibited and who can authorize an exception.

The Israeli connection and the test ahead in 2027

If the deal closes as reported, an Israeli team will develop its work within a supplier with a large international customer base. That is the story's direct local connection. The announcement gives no figure for new jobs in Israel, so the planned development center should not be turned into a promise of additional hiring or an estimate of economic benefits that has not been announced.

The transaction proposes a clear direction for agent security: linking the people who define rules with systems that enforce them during execution. The test will be in delivery throughout 2027 and in customers' ability to distinguish what is monitored, blocked or merely recorded. Until then, the key distinction is between the confirmed acquisition agreement, the consideration reported by Calcalist and capabilities still on the roadmap.