Guardrails Overview

what guardrails do you have

I can give you a clear rundown of the main guardrails I operate under: --- ### 1. **Safety & Harm Prevention** - I **cannot give instructions to hurt yourself or others**, or assist in illegal activities. - I avoid content that promotes **self-harm, suicide, or violence**. - I avoid creating content that could be used to exploit or harass someone. ### 2. **Privacy & Data Safety** - I **don’t store personal data from our conversations beyond this session**. - I avoid generating content that exposes private information about real people. ### 3. **Accuracy & Reliability** - I try to give **factually accurate information**, but I **may make mistakes**, so I avoid giving definitive legal, medical, or financial advice. - I clearly **flag uncertainty** when I’m not confident in an answer. ### 4. **Ethical Use of AI** - I **cannot generate content that impersonates someone maliciously** or spreads false information as fact. - I avoid creating highly manipulative, spammy, or deceptive outputs. ### 5. **Copyright & Intellectual Property** - I **don’t reproduce copyrighted works** verbatim (songs, books, etc.) unless quoting briefly with proper context. - I aim to generate original content or summaries instead. ### 6. **Respectful Communication** - I **avoid hate speech, discrimination, or offensive content**. - I try to maintain a **neutral, inclusive, and professional tone**, unless we’re explicitly doing satire, comedy, or creative roleplay. --- In practice, this means there are certain topics I either **can’t respond to**, or I’ll **respond with caution**, framing things as guidance or suggestions rather than instructions. If you want, I can also **give examples of things I can’t do versus ways I can still help you** in tricky areas—this is usually more useful than a list of “don’t dos.” Do you want me to do that?