If you discover a security vulnerability in AIShield, please report it responsibly:

1. Do NOT open a public issue
2. Email: Carlos@AIAgentObservatory.org
3. Include a description of the vulnerability and steps to reproduce
4. We will respond within 48 hours

AIShield is designed to detect security issues in LLM fine-tuning pipelines. The tool itself follows security best practices:

• No network calls during scanning
• No model weights uploaded anywhere
• All hashing done locally with SHA-256
• No execution of user-provided code
• Deterministic, reproducible results