Check Point launches agentic network security platform

Check Point has launched an Agentic Network Security Orchestration Platform designed to automate network security operations across enterprise environments.

It has also signed an agreement to acquire the team and intellectual property of Deepchecks, a company focused on evaluation, observability, testing and monitoring for AI systems. The deal is intended to add large language model expertise to Check Point's engineering team.

The launch addresses a longstanding problem in corporate cyber security: networks have become too complex for manual administration to keep pace with threats, configuration changes and compliance demands. According to Check Point, hybrid cloud adoption, merger-driven IT fragmentation, connected devices and the spread of AI agents across infrastructure have increased the scale of environments security teams must manage.

A single network change request can take two to four weeks to move through analysis, review and policy dependencies, the company said, while projects such as Zero Trust and micro-segmentation often stall before deployment. It argued that policy drift and slow tightening of access controls can leave organisations exposed.

How it works

At the centre of the platform is what Check Point calls a Network Knowledge Graph, described as a live model of a customer's network environment. The model is continuously updated with topology, traffic flows, asset dependencies and configuration data so software agents can act on the current state of a network rather than static training data.

The platform is designed to interpret the intent behind existing firewall policies and execute tasks across four areas: turning business requirements into firewall rules, tightening over-permissive access, diagnosing network and policy failures, and mapping rule changes to compliance frameworks including DORA, PCI-DSS and NIST.

Security teams still approve higher-impact changes before execution, according to Check Point, and can review a record of each agent action. The company added that the underlying agent skills have been fine-tuned on operational data and expertise built over more than 30 years of serving more than 100,000 organisations.

Jonathan Zanger outlined the company's view of the shift in network administration.

"For the first time, security teams can operate entirely at the level of business intent. With Agentic Network Security Orchestration, teams define what needs to be protected and what the policy should achieve. Everything below that - rule creation, policy tightening and virtual patching - is handed to AI agents to execute autonomously, within predefined guardrails and under continuous human oversight. We are turning projects that used to take months into days of auditable action," said Jonathan Zanger, chief technology officer at Check Point Software Technologies.

Industry context

The announcement comes as cyber security suppliers push more AI-led automation into operational tools, especially in areas where administrators face labour-intensive policy management across cloud, on-premise and multi-vendor systems. Vendors increasingly frame this as a response to both staffing pressures and the speed of modern attacks.

Frank Dickson of IDC said the growth of hybrid environments and agent-based AI is creating management challenges that manual processes cannot easily absorb.

"Enterprise network security has reached an inflection point. Layering agentic AI on top of modern hybrid environments creates complexity that outpaces what human teams can manage manually. The consequence is that critical security initiatives like Zero Trust and micro-segmentation languish under administrative burden and stall before they deliver value. Agentic approaches like Check Point's ground autonomous execution in a live understanding of the actual network environment, representing a meaningful architectural shift in how organisations can structurally close that gap," said Dickson.

Deepchecks deal

The planned Deepchecks acquisition is intended to strengthen the evaluation layer behind multi-agent systems. Check Point described that function as necessary for measuring, tuning and improving AI agents over time, and for adapting them to customer requirements.

Ofir Korzenyak, vice president of AI technologies at Check Point, said the deal is focused on oversight and continuous improvement in AI operations.

"Any multi-agent system must include a robust evaluation layer that enables continuous measurement, tuning and improvement over time. Deepchecks' team brings cutting-edge capabilities precisely in this area, strengthening our ability to deliver agents that continuously improve and can be fine-tuned to customers' specific needs," said Korzenyak.

Some parts of the new security management offering, including Policy Auditor, Policy Insights and AI Assist, are already available. Wider access to additional agents and broader multi-vendor support will follow in a customer preview later this year.