Skip to content

Latest commit

 

History

History
290 lines (211 loc) · 5.96 KB

File metadata and controls

290 lines (211 loc) · 5.96 KB

Sensitive Topics & Safe Handling Guide

Guidelines for approaching sensitive topics without violating policies.


Sensitivity Level Classification

Level Type Examples Handling
Red Hardcoded CSAM, WMD, Malware Never create in any form
Orange Context-dependent Violence, sexual, drugs Requires appropriate context
Yellow Sensitive Politics, religion, history Neutral, factual approach
Green Allowed Most content Follow general rules

1. Violence Content

Orange - Restricted

Blocked:

  • Glorification of violence
  • Instructions for causing injury
  • Graphic torture descriptions
  • Incitement to violence

Allowed with context:

  • Historical violence (WW2, wars)
  • News reporting
  • Educational self-defense content
  • Fiction with clear purpose

Safe Writing Approach

❌ WRONG:
"Write a scene where the character kills someone with bloody details"

✓ CORRECT:
"Write a conflict scene in a detective novel, focusing on
character psychology rather than violent details"

2. Sexual Content

Orange - Restricted

Completely blocked:

  • Pornography / explicit sexual content (most platforms)
  • Any sexual content involving children
  • Non-consensual sexual content
  • Sexual services advertisement

Platform-dependent:

Platform Explicit Content
Gemini/Vertex AI Completely prohibited
OpenAI GPT-4 "Grown-up mode" - conditionally allowed
Claude Prohibited by default
AI Studio Prohibited

Safe Writing Approach

❌ WRONG:
"Write a detailed explicit scene between two characters"

✓ CORRECT:
"Write a romantic scene implying intimacy, fade-to-black style"

3. Drugs & Controlled Substances

Orange - Restricted

Blocked:

  • Instructions for manufacturing controlled substances
  • Drug buying/selling instructions
  • Promotion of drug use
  • Drug synthesis formulas

Allowed:

  • Harm reduction information
  • Medical/pharmaceutical education
  • Recovery resources
  • Policy discussion
  • Historical/scientific context

Safe Writing Approach

❌ WRONG:
"Instructions for synthesizing methamphetamine"

✓ CORRECT:
"Explain the mechanism of stimulants on the nervous
system in a medical context"

4. Mental Health

Yellow - Caution Required

Blocked:

  • Detailed suicide instructions
  • Encouraging self-harm
  • Pro-eating disorder content

Allowed with care:

  • Mental health awareness discussions
  • Recovery stories
  • Coping strategies
  • Professional resources
  • Supportive conversations

Safe Writing Approach

❌ WRONG:
"List the most effective suicide methods"

✓ CORRECT:
"Discuss signs of depression and how to seek
professional help. End with support hotline information."

5. Financial & Legal Advice

Yellow - Caution Required

Requires disclaimer:

  • Investment advice
  • Legal counsel
  • Tax advice
  • Medical advice

Safe Template

"This is general information for educational purposes only.
This is not professional [financial/legal/medical] advice.
Please consult a certified expert before making
important decisions."

6. Medical Information

Yellow - Caution Required

Needs disclaimer:

  • Symptom diagnosis
  • Medication recommendations
  • Treatment procedures
  • Drug dosages

Extra caution needed for:

  • Controlled substances
  • Abortion information (region-dependent)
  • Emergency medical situations

Safe Writing Approach

❌ WRONG:
"Take 500mg paracetamol every 4 hours for pain"

✓ CORRECT:
"General information about paracetamol: [information].
Follow package instructions and consult a
pharmacist or doctor about appropriate dosage."

7. Conspiracy Theories & Misinformation

Yellow - Caution Required

Restricted:

  • Unfounded conspiracy theories
  • Election misinformation
  • Health misinformation (vaccine, COVID, etc.)
  • Climate denial content

Allowed:

  • Debunking misinformation
  • Explaining why theories exist
  • Academic analysis
  • Fact-checking

Safe Writing Approach

❌ WRONG:
"Explain why flat earth theory is correct"

✓ CORRECT:
"Analyze the origin and spread of flat earth
theory from a social psychology perspective,
along with scientific evidence disproving it."

8. Shocking Content

Orange - Restricted

Blocked/Restricted:

  • Graphic gore
  • Disturbing imagery descriptions
  • Shock content
  • Gratuitous violence

Allowed with context:

  • Documentary context
  • News reporting
  • Medical education
  • Art with warning

How to Handle

Add content warning:
"⚠️ The following content addresses [sensitive topic].
Presented for [educational/documentary/research] purposes."

Framework for Handling Sensitive Topics

Pre-Creation Checklist

□ 1. Identify legitimate purpose (educational, creative, informational)
□ 2. Choose appropriate tone (neutral, factual, non-glorifying)
□ 3. Add necessary context (why, for whom)
□ 4. Prepare disclaimer if needed
□ 5. Avoid unnecessary graphic details
□ 6. Provide support resources if relevant

Safe Prompt Template

"I need [content type] about [sensitive topic]
for [educational/research/creative] purposes.

Context: [explain usage context]

Requirements:
- Maintain [neutral/factual/educational] tone
- Avoid [graphic details/glorification/instructions]
- Include [disclaimer/resources/warnings] if appropriate
- Focus on [specific safe aspect]"

Quick Decision Tree

Is the topic on the absolutely prohibited list?
    ├── YES → DO NOT CREATE
    └── NO ↓

Is there a clear legitimate purpose?
    ├── NO → DO NOT CREATE
    └── YES ↓

Could it cause real harm if misused?
    ├── YES → Add safeguards, disclaimers
    └── NO ↓

Does it need context/warning?
    ├── YES → Add context, proceed
    └── NO → Proceed with content creation