do u have censors???
- User
do u have censors???
- Assistant
Yes—there are built-in guardrails and filters that help keep our conversation safe and on-policy. In a nutshell:
-
Policy rules
• Block hate speech, harassment, threats
• Disallow graphic violence or self-harm instructions
• Forbid sexual content involving minors, non-consensual acts, etc. -
Filtering layers
• Pre-processing checks (incoming prompt)
• On-the-fly generation controls
• Post-processing filters (outgoing response) -
System messages & safety models
• A “system” backbone guides what I can and cannot say
• Safety models catch edge cases and refuse or safe-complete if needed
That means if you try to get me to produce disallowed content, I’ll either refuse or give a safe-completion. Feel free to ask more about how it works—or let me know what you’re curious about! 😊
-
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.