Loading page…
The new Falcon system combines CrowdStrike’s security data and agent harnesses with NVIDIA’s open Nemotron models. Early tests favor the optimized open pipeline on detection quality, but cover only one attack scenario family.
Listen to this story
CrowdStrike’s SafeMind turns detection engineering into an adversarial simulation loop: red agents attack a simulated NVIDIA accelerated-computing environment, while blue agents generate and validate Falcon detections. Its optimized Nemotron-based pipeline lifted mean detection from 16.5% to 41.9% in limited backtests, but the comparison changed models and harnesses together. Notably, three open-pipeline detections earned “gold” status and covered all eight attacks, versus none for the frontier system.
Nemotron 3 Ultra orchestrates defense; fine-tuned Nemotron 3 Super writes detection rules.
The optimized pipeline produced 11 detections; five fired on unseen attacks, while only three passed independent gold review.
The frontier baseline generated 35 backtest-passing detections, but 10 fired on unseen attacks and none qualified as gold.
CrowdStrike has introduced SafeMind inside its Falcon platform, pairing its cybersecurity models and agent harnesses with NVIDIA Nemotron open models. The system is being tested in a simulated version of NVIDIA’s accelerated-computing infrastructure, where an attacking agent and a defending agent repeatedly try to outmaneuver each other. The premise is to turn a slow, manual detection-engineering cycle into a machine-speed loop.
SafeMind’s division of labor is deliberate. Nemotron 3 Ultra orchestrates the defensive harness, while a fine-tuned Nemotron 3 Super writes detection rules. CrowdStrike post-trained the models on its cybersecurity experience and threat data, then paired them with proprietary harnesses: the software layer that gives a language model tools, context and a repeatable workflow.
That workflow begins with a threat-informed objective. A red-team harness selects and executes an attack path in the isolated environment, while Falcon sensors capture its traces. The blue-team harness uses those traces and telemetry to identify gaps, generate candidate detections and send them through checks before deployment. A new, independently seeded attack then tests whether the detection holds up or the red agent can find an alternate route.
The live-fire comparison complicates any simple scorecard. The complete frontier system produced 35 backtest-passing detections, versus 11 from the optimized open pipeline. But five of the 11 open detections fired against at least one of eight unseen attacks, compared with 10 of 35 frontier detections. Four open and nine frontier detections stayed quiet on available test traffic; after independent review, three open detections qualified as gold and none of the frontier detections did.
That outcome supports a narrower claim than a general open-versus-proprietary verdict. The optimized pipeline’s three gold detections covered all eight attacks, while the frontier system’s detections covered seven. Yet the study covered one scenario family and small detection sets, leaving performance across other environments and attacks untested.
CrowdStrike also introduced Falcon IQ, a separate agentic workload-automation system for assessment, prioritization and remediation. It uses more than 50 agents as a unified workforce and runs through Charlotte AI AgentWorks, CrowdStrike’s no-code platform whose agentic engine is helped by NVIDIA Nemotron models. The distinction matters: SafeMind is centered on adversarial testing and detection creation; Falcon IQ is aimed at operational security workflows.
CrowdStrike separately says its Blue Solano model, based on Nemotron 3 Super, exceeded leading frontier models’ accuracy at 99% lower cost in internal evaluations. The disclosed controlled test is more specific: it shows a validation-heavy open pipeline outperforming a frontier system on selected measures, while leaving the broader generalization question open.
Editorial analysis
SafeMind is a test of whether open models become more valuable in security when paired with proprietary telemetry, task-specific training and rigorous validation. NVIDIA’s role reaches beyond supplying a base model: its accelerated-computing infrastructure was modeled in the representative environment where the agents were evaluated. The consequential result is not CrowdStrike’s broad accuracy-and-cost claim, but the smaller live-fire comparison: three open-pipeline detections qualified as gold and covered all eight attacks in one scenario family. The next evidence to watch is whether that result repeats across different environments and attack families.
Loading discussion...
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.