CrowdStrike Accelerates Real-Time Data Classification with On-Device AI

CrowdStrike describes a new Falcon Data Security capability that uses on-device language models and Intel neural processing units to classify sensitive structured and unstructured data in real time, including credentials, documents, chat messages, AI prompts, and medical records. The approach aims to improve on pattern-based detection by interpreting contextual meaning while avoiding cloud-inference latency and privacy concerns associated with transmitting customer data. Developed with Intel for Core Ultra AI PCs, the system uses NPU-optimized models through ONNX Runtime and OpenVINO, selecting NPUs over GPUs for sustained, low-power background processing despite GPUs’ faster raw inference. CrowdStrike trained a large model and distilled it into a smaller FP16-quantized endpoint model that uses under 5% of the teacher model’s parameters while retaining similar classification effectiveness. The underlying portable ONNX architecture is intended to support different hardware environments, including NVIDIA GPU cloud deployments and Apple Neural Engine-based macOS devices, allowing the company to extend context-aware data protection across endpoints and cloud platforms.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.