TrendAI Extends NVIDIA Agent Security With Threat Intelligence
TrendAI, a provider of AI-powered enterprise cybersecurity software, has announced support for the NVIDIA Agent Safety Platform, combining its Vision One security technology with NVIDIA infrastructure designed to protect autonomous AI agents.
TrendAI Vision One provides inline inspection of agent prompts and responses to detect prompt injection and jailbreak attempts and prevent sensitive information from leaking. It also lets security teams govern which tools and Model Context Protocol servers agents can access and inspect their tool calls.
Data and identity controls discover and classify information used by AI systems, apply data loss prevention to files agents access, and address excessive privileges assigned to agents and other non-human identities.
The integration also uses telemetry from NVIDIA BlueField data processing units and NVIDIA DOCA. TrendAI says this allows detection to operate independently of the host operating system and helps correlate AI activity with endpoint, network and identity telemetry.
NVIDIA's Agent Safety Platform provides infrastructure-level policy enforcement, including OpenShell, which applies controls outside an agent's execution environment.
For enterprise security teams, the approach reflects the expanding scope of agentic AI security. Protecting production agents increasingly requires controls extending beyond model inputs and outputs to identities, data, tools, networks and the underlying compute infrastructure.
Posted by Pure AI Editors on 09/28/2026