CrowdStrike and NVIDIA are expanding their collaboration to protect autonomous AI agents across both software and computing infrastructure.
The planned integration is intended to combine agent prompts, model interactions and runtime activity with CrowdStrike’s wider security telemetry. That would allow organizations to identify suspicious behavior across agent workflows while preserving policy controls outside the agent process. CrowdStrike said the collaboration will also extend to NVIDIA Sentry on BlueField-4 data-processing units, where monitoring and policy enforcement operate independently from the software runtime.
NVIDIA describes Sentry as an out-of-band watchdog that uses its DOCA software to inspect agent requests and responses, verify agent identity, provide attested telemetry and enforce zero-trust access rules. If an agent attempts to leave its permitted boundary, the system is designed to quarantine it in milliseconds. OpenShell is open source and can be extended to run on computing platforms from Arm and Intel, while Sentry provides the additional hardware-isolated controls available through BlueField systems.
The announcement follows NVIDIA’s September launch of its Open Agent Safety Platform, which combines OpenShell’s runtime governance with Sentry’s silicon-level monitoring. NVIDIA said more than 100 organizations were using the platform at launch, including Microsoft, Perplexity, Accenture and JPMorgan Chase. CrowdStrike’s hardware integration remains described as planned, so the announcement establishes a development direction rather than a completed silicon-enforcement product.