AI you can inspect, running on your own infrastructure. Every verdict is explainable and verified against the underlying evidence, and detection keeps working fully offline.
Does GRYPHON use artificial intelligence?
Yes, and it does real work. GRYPHON pairs inspectable rules with a purpose-built transformer model for behavioral detection, so you keep full visibility into why something was flagged. Everything runs on-premises, is explainable, and needs no cloud. No black boxes, no data leaving your environment.
Behavioral detection AI
- A purpose-built transformer model learns how software behaves and flags malicious activity.
- It watches what a program does across processes, files, the registry and the network, not just known signatures.
- A final decision layer turns that into a clear verdict, benign or malicious, tuned to keep false alarms low.
- Detects ransomware, trojans and code injection, each with its own sensitivity settings.
- Explainable by design: every verdict shows what triggered it.
Trained by us, runs on your endpoints
- Detection models are trained and tuned on IstroSec’s own compute infrastructure, not on yours.
- They export to a self-contained format for efficient on-device inference, so protection keeps working even fully offline/air-gapped.
- No cloud dependency and no special hardware: detection, blocking and recovery run locally on standard Windows endpoints.