"What they have been building over the past five years or so is really a platform approach and expertise in evaluating how good models are," Chief Technology Officer Jonathan Zanger told ISMG.
Check Point’s strategic purchase: Deepchecks to validate AI-driven actions
Check Point Software has moved to acquire Tel Aviv-based Deepchecks to bring model- and agent-validation capabilities into its network security orchestration, the company said. The Silicon Valley–based platform security vendor described the proposed acquisition as a way to help organizations test, evaluate and monitor machine learning systems and agentic workflows — capabilities the firm calls increasingly essential as generative AI introduces operational risks including hallucinations.
Jonathan Zanger, Check Point’s chief technology officer, framed the deal as a response to the need for measurable confidence in autonomous systems that will operate inside enterprise networks. "How can you make sure that the agents are actually behaving correctly?" he asked. Zanger said the acquisition will focus on ensuring agents operate inside the right guardrails when dealing with mission-critical tasks.
Deepchecks: small team, focused IP, and a military-linked founder
Deepchecks was founded in 2019 and, as of the announcement, employs 15 people. The company closed a $14 million seed funding round in June 2023 led by Grove Ventures. Deepchecks’ founder and long-time leader, Philip Tannor, previously spent nearly seven years in the Israel Defense Forces, including two-and-a-half years supervising research in computer vision, natural language processing and signal processing — experience Check Point emphasized when describing the startup’s talent and expertise.
Validating agentic workflows where correct answers can vary
Check Point and Deepchecks argue that traditional validation approaches are poorly matched to agentic generative-AI workflows, where correct outputs can be inconsistent across interactions. "When we are dealing with GenAI, the answer can change but still be correct," Zanger said. Deepchecks has developed intellectual property aimed specifically at evaluating these agentic workflows — measuring whether a system’s reasoning and outcome meet correctness criteria even if the surface outputs differ.
That validation focus extends to ensuring agents consult trusted internal data sources, ask the right questions, and consistently generate accurate operational guidance. Zanger framed this capability as essential for identifying an enterprise’s "crown jewels" and preventing agents from hallucinating or behaving unpredictably during high-stakes security operations.
Use cases named: microsegmentation, threat prevention, threat intelligence
Check Point described concrete operational areas where embedded validation could make a difference. The company said its new platform will let autonomous agents assist with configuration, analysis and operational tasks while humans retain oversight; one explicit example Zanger offered was microsegmentation. "When CISOs ask agents to handle major projects like microsegmentation, we can actually make sure that those agents are behaving as well as expected," he said.
Zanger also listed threat prevention, threat intelligence and AI security capabilities as initial application areas for Deepchecks’ validation layer, and said Check Point ultimately intends to embed the technology throughout its portfolio to provide an additional layer of trust and optimization.
What this means for security administrators, CISOs, and procurement leaders
- Security administrators and CISOs: They will get validation tools intended to verify agent behavior in operational tasks and to ensure systems either deliver correct results or explicitly acknowledge uncertainty, reducing the risk of fabricated responses in mission-critical environments.
- Procurement leaders and enterprise buyers: The acquisition gives them a vendor claim to measurable validation and monitoring for agent-driven workflows — a selling point Check Point sees as central to convincing practitioners to adopt AI-enabled orchestration.
- Technologists and data scientists: Deepchecks’ platform and IP aim to provide evaluative methods for correctness where outputs vary, supporting teams that must judge contextual correctness across inconsistent but valid responses.
Embedding validation and the promise of trust — a concrete next step
Check Point said it plans to fold Deepchecks’ validation technology into its network security orchestration and broader product lines. Zanger emphasized trust as the core enabler of adoption: "We believe that trust is a very big part of the willingness of security practitioners to adopt this kind of technology, and the ability to show them how accurate actually their agents are will boost the confidence in the ability to adopt this new and disruptive technology."
The deal puts a focused validation capability — a 15-person team, IP tuned for agentic correctness, and the founder’s research background — at the center of Check Point’s push to make AI agents operationally reliable. Whether embedding that capability across the portfolio measurably reduces hallucinations in mission-critical security operations and accelerates practitioner adoption remains the key implementation step Check Point has signaled it will pursue.
Source: Check Point Validates AI-Driven Actions With Deepchecks Buy — GovInfoSecurity
