[Bug] Overly aggressive security safeguard triggering on benign security/cyber terminology

Bug Description
Keeps falling back into the security safeguards, and downgrades the model, which is rather annoying, it almost seems that when it sees the word cyber, or security, it will immediatly be triggered; there was no malicious thing going on here, just me trying to improve my product.

Environment Info

• Platform: darwin
• Terminal: zed
• Version: 2.1.222
• Feedback ID: a6e14a2e-03e9-4686-a32e-63cc86417e04

Errors

[]