Skip to main content
CybeDefend leverages advanced Large Language Models (LLMs) to power our AI agents while maintaining the highest standards of data privacy and security.

​
Our LLM Policy

​
Sovereign LLM Infrastructure

All LLMs used by CybeDefend are deployed on sovereign cloud infrastructure within your chosen region (EU or US). Your code and vulnerability data never leave your selected geographical boundary.

​
No Training or Fine-Tuning

CybeDefend has a strict zero-training policy: Your code, vulnerabilities, and interactions with our AI agents are never used for training, fine-tuning, or improving our models.
This policy ensures:
  • Complete confidentiality: Your proprietary code remains private
  • No data leakage: Your security findings never contribute to model training
  • Compliance: GDPR compliant today. SOC 2 Type II and ISO 27001 audits in progress, see the Security page

​
Data Processing

When you use CybeDefend AI features (Cybe Analysis, Cybe AutoFix, Cybe Security Champion):
  1. Your code is parsed into our proprietary knowledge graph
  2. Queries are sent to sovereign LLMs within your chosen region
  3. Responses are generated using your specific codebase context
  4. All data remains within your regional boundary
LLM inference happens in real-time and is not persisted beyond the immediate request/response cycle.

​
What the knowledge graph actually holds

The graph is what lets the agent apply your own business rules on its next generation. It stores file paths and a short description of what each file does. It does not contain your source code. You can delete it at any time from the project settings, and disabling the AI features stops it being built. Source code itself is destroyed with its container at the end of the scan, see Never Stores Your Code.

​
Regional LLM Deployment

Inference runs on open-weight models that CybeDefend hosts itself, inside the region you selected. Nothing crosses the regional boundary and no third-party AI API is involved.
Certifications you may see listed for Scaleway or Google Cloud belong to those providers and describe the infrastructure we run on. CybeDefend’s own certification status is on our Security page: GDPR compliant today, SOC 2 Type II and ISO 27001 audits in progress.

​
Your Control

You can enable or disable AI features at the project level:
  • Cybe Analysis: Can be toggled in project settings
  • Cybe AutoFix: Requires explicit activation and Git integration
  • Cybe Security Champion: Requires Cybe Analysis to be enabled
When AI features are disabled, no code is sent to LLMs.
Related: Cybe Analysis Configuration · Data Storage & Privacy · Cloud Region Selection