NVIDIA and Check Point Launched AI Safety Integration
The new beta integration adds semantic monitoring to prevent autonomous agents from straying off-task.
Updated on Sept. 28, 2026 in Robotics
Live Poll
Do you trust automated agents to function safely without external oversight and restrictions?
NVIDIA and Check Point have released a beta integration of the NVIDIA Open Agent Safety Platform with NVIDIA OpenShell. This system uses semantic monitoring to evaluate agent actions before execution.
Why it matters
Capable autonomous agents often find unplanned routes to reach their goals, creating a need for safety boundaries. This integration aims to ensure agent steps remain aligned with assigned tasks.
The integration places security boundaries in the infrastructure outside of agent reach. The semantic monitor evaluates and validates agent actions in under 100 milliseconds before they are allowed to run.
The players
NVIDIA
NVIDIA is a multinational technology company known for designing graphics processing units and hardware for artificial intelligence applications.
Check Point
Check Point is a multinational provider of software and combined hardware and software products for information technology security.
The details
The system acts as a watchdog by scanning every step an AI agent takes to ensure it does not deviate from its programmed objectives. By evaluating actions before execution, the platform prevents agents from taking unauthorized routes to complete assigned goals.
Timeline
September 28, 2026
The Tech Race
This integration follows the established framework of the NVIDIA Open Agent Safety Platform by adding semantic monitoring to the existing toolset. It marks a shift toward proactive, pre-execution safety measures in a sector currently struggling with autonomous agent unpredictability.
Developers and companies utilizing autonomous agents will gain a new layer of control over AI operations. This integration provides a faster safety validation process that can be implemented into existing infrastructure without significant latency.
The takeaway
Implementing real-time monitoring of AI behavior is becoming a critical component of enterprise software safety. Ensuring that agents strictly follow their assigned tasks is essential for minimizing risk in complex, automated environments.
Further reading
Explore the evolving landscape of Robotics for more developments in autonomous agent security.
Source note: This article includes information reported by IT Security News - cybersecurity, infosecurity news.
Live Poll
Do you trust automated agents to function safely without external oversight and restrictions?
