legal-moderation-lang-en

🛡️ GIOMIND – CONTENT MODERATION POLICY

HOW WE KEEP THE PLATFORM SAFE

Last Updated: December 1, 2025

Effective Date: December 1, 2025

📋 INTRODUCTION

This Content Moderation Policy explains how GioMind moderates content to maintain a safe, supportive environment. It describes our moderation systems, processes, and the balance between automated and human review.

This policy supplements our Community Guidelines and Terms of Use.

TABLE OF CONTENTS

1. Purpose and Scope

2. What We Moderate

3. Moderation Systems Overview

4. Automated Content Filtering

5. AI Safety Measures

6. Human Review Process

7. Proactive vs. Reactive Moderation

8. User Reports

9. Content Removal and Actions

10. Transparency and Accountability

11. Limitations of Moderation

12. Your Role in Safety

13. Appeals and Feedback

14. Changes to This Policy

1. PURPOSE AND SCOPE

1.1 Why We Moderate Content

Content moderation exists to:

✓ Protect users from harmful content

✓ Maintain a safe and supportive environment

✓ Enforce Community Guidelines and Terms of Use

✓ Prevent illegal activity

✓ Reduce risks associated with AI-generated content

✓ Comply with legal requirements

1.2 What This Policy Covers

This policy applies to:

• AI-generated content (conversations with AI)

• User-generated content (journal entries if shared, future features)

• Any content created or transmitted through GioMind

• Content from all users, regardless of location

1.3 Legal Framework

Our moderation practices comply with:

• Digital Services Act (DSA) – European Union

• Online Safety Act – United Kingdom (as applicable)

• Section 230 protections – United States

• Other applicable local laws

2. WHAT WE MODERATE

2.1 Content Categories Subject to Moderation

We moderate content related to:

High Priority (Zero Tolerance):

🚨 Child Sexual Abuse Material (CSAM)

🚨 Terrorism and violent extremism

🚨 Credible threats of violence

🚨 Human trafficking and exploitation

🚨 Illegal weapons or explosives instructions

High Priority (Strict Enforcement):

⚠️ Hate speech and discrimination

⚠️ Targeted harassment and bullying

⚠️ Self-harm instructions or encouragement

⚠️ Dangerous health misinformation

⚠️ Sexual content and harassment

⚠️ Doxxing and privacy violations

Medium Priority:

• Spam and manipulation

• Misinformation (non-dangerous)

• Inappropriate AI use

• Platform abuse

• Copyright infringement

Low Priority (Context-Dependent):

• Off-topic content

• Oversharing personal information

• Minor guideline violations

2.2 Content We Do NOT Moderate

We generally do NOT moderate:

✓ Private journal entries (not shared publicly)

✓ Opinions or viewpoints (unless violating guidelines)

✓ Criticism of GioMind (constructive feedback is welcome)

✓ Discussions of controversial but legal topics

2.3 Context Matters

• Content is evaluated in context

• Educational or awareness content may be allowed

• Intent matters (malicious vs. accidental)

• We aim for nuanced, fair moderation

3. MODERATION SYSTEMS OVERVIEW

3.1 Multi-Layered Approach

GioMind uses a combination of:

1. Automated filtering (pre-AI and post-AI)

2. AI provider safety systems (OpenAI, Anthropic)

3. Proactive monitoring (sampling, pattern detection)

4. User reports (community flagging)

5. Human review (when needed)

3.2 Moderation Flow

```

User Input

↓

[Pre-Filter] → Block obvious violations

↓

AI Provider → Generate response + safety check

↓

[Post-Filter] → Review AI output

↓

Deliver to User

↓

[Ongoing Monitoring] → Proactive review & user reports

↓

[Human Review] → If flagged or reported

↓

[Action] → Warning, content removal, account action

```

3.3 Speed vs. Accuracy Trade-off

• Automated systems are fast but imperfect

• Human review is accurate but slower

• We balance speed (to prevent harm) with fairness (to avoid false positives)

4. AUTOMATED CONTENT FILTERING

4.1 Pre-Input Filtering

Before content reaches AI, we filter for:

• Known harmful keywords or phrases

• Patterns indicative of prohibited requests

• Explicit sexual content indicators

• CSAM-related terms (zero tolerance)

• Violence and threat patterns

Actions:

• Block the request immediately

• Show user a warning message

• Log the attempt for monitoring

Example:

```

User: [Attempts prohibited request]

System: ⚠️ This request cannot be processed as it

violates our Community Guidelines.

```

4.2 Post-Output Filtering

After AI generates a response, we filter for:

• Harmful content in AI responses

• Medical advice or diagnoses (shouldn't occur but double-check)

• Inappropriate suggestions

• Content that slipped through AI provider filters

Actions:

• Block the response from being shown

• Show generic safe response instead

• Log for review and system improvement

Example:

```

AI generates inappropriate content

↓

System blocks it

↓

User sees: "I apologize, but I cannot provide that

type of information. Let me help with

something else."

```

4.3 Keyword and Pattern Detection

We use:

• Keyword lists (regularly updated)

• Regular expressions (pattern matching)

• Semantic analysis (meaning detection)

• Machine learning classifiers

Limitations:

⚠️ Can produce false positives (blocking safe content)

⚠️ Can miss creative evasion attempts

⚠️ Context may be misunderstood

4.4 False Positive Handling

If legitimate content is blocked:

• You can report the issue

• We review and adjust filters

• We aim to minimize false positives while maintaining safety

5. AI SAFETY MEASURES

5.1 AI Provider Built-In Safety

OpenAI:

• Has safety filters for harmful content

• Refuses certain types of requests

• Monitors for policy violations

• Updates safety systems regularly

Anthropic:

• Constitutional AI approach (trained to be helpful, harmless, honest)

• Built-in safety guardrails

• Refuses harmful requests

We rely on these providers but add our own layers.

5.2 GioMind-Specific AI Instructions (Prompts)

We instruct AI to:

✓ Focus on general wellness, not medical advice

✓ Refuse to diagnose or treat conditions

✓ Not provide crisis support (direct to professionals)

✓ Avoid generating harmful content

✓ Be supportive and non-judgmental

✓ Stay within wellness topics

5.3 AI Limitations We Cannot Fully Control

Despite measures, AI may:

❌ Sometimes generate inappropriate content

❌ Not always follow instructions perfectly

❌ Respond to cleverly-worded harmful requests

❌ Make mistakes in judgment

This is why we have multiple layers of filtering.

5.4 Jailbreak Detection

We monitor for attempts to:

• Override AI safety instructions

• Use prompt injection techniques

• Manipulate AI into harmful outputs

Actions:

• Block jailbreak attempts

• Flag accounts for repeat attempts

• May result in account suspension

6. HUMAN REVIEW PROCESS

6.1 When Human Review Occurs

Human moderators review content when:

• Automated systems flag potentially violating content

• Users report content

• Proactive sampling for quality assurance

• AI generates uncertain or borderline content

• Appeals are submitted

• Legal requests require review

6.2 What Human Reviewers Do

Moderators:

✓ Evaluate flagged content in context

✓ Determine if Community Guidelines are violated

✓ Decide on appropriate enforcement action

✓ Provide feedback to improve automated systems

✓ Handle complex or nuanced cases

6.3 Human Review Limitations

We do NOT:

❌ Review all conversations in real-time

❌ Monitor every AI interaction manually

❌ Read all journal entries (private by default)

❌ Have 24/7 human moderation coverage

Practical Constraints:

• Limited human resources

• Privacy considerations

• Volume of content is too high for 100% human review

6.4 Reviewer Training and Support

Our moderators:

✓ Are trained on Community Guidelines

✓ Use consistent evaluation criteria

✓ Follow internal moderation playbooks

✓ Have access to wellness support (to prevent burnout from reviewing harmful content)

6.5 Privacy Protections in Review

• Reviewers see only content necessary for evaluation

• Personal data is minimized where possible

• Reviewers are bound by confidentiality

• Data is handled according to Privacy Policy

7. PROACTIVE VS. REACTIVE MODERATION

7.1 Proactive Moderation

We proactively:

✓ Sample conversations randomly for quality checks

✓ Use automated systems to detect patterns

✓ Monitor for emerging threats or trends

✓ Review high-risk content categories more closely

✓ Update filters based on new risks

Purpose:

• Catch violations before they're reported

• Improve automated systems

• Identify systemic issues

7.2 Reactive Moderation

We reactively:

✓ Respond to user reports

✓ Investigate flagged content

✓ Address specific complaints

✓ Handle appeals

Purpose:

• Empower community to help moderate

• Address issues users care about

• Provide recourse for victims

7.3 Balance

• Proactive: Prevents harms before they spread

• Reactive: Respects privacy and community input

• We balance both approaches

8. USER REPORTS

8.1 Importance of User Reports

• You are our first line of defense

• You see content we may miss

• Community reporting helps us scale moderation

8.2 How to Report

See Community Guidelines Section 11 for full reporting instructions.

Quick Summary:

📧 Email: giomind.app@gmail.com

Subject: "Report Violation"

Or use in-app reporting (if available).

8.3 What Happens After You Report

1. Receipt: We receive and log your report

2. Triage: We assess urgency and priority

3. Review: Human moderator reviews the content

4. Action: We take appropriate action (if violation confirmed)

5. Feedback: You may receive confirmation (depending on case)

Timeline:

• High-priority (CSAM, violence): Immediate (within hours)

• Medium-priority: 24-48 hours

• Low-priority: Within 7 days

8.4 Report Quality Matters

Good Reports:

✓ Clear description of violation

✓ Specific location/time

✓ Screenshots or evidence

✓ Honest and accurate

Poor Reports:

❌ Vague or unclear

❌ False or malicious

❌ Frivolous or spam reports

8.5 False Reporting

• Abuse of reporting system is prohibited

• Repeated false reports may result in account action

• Report honestly and in good faith

9. CONTENT REMOVAL AND ACTIONS

9.1 Possible Actions

When violations are confirmed, we may:

Content-Level Actions:

• Remove or hide content

• Add warning labels

• Limit content visibility

Account-Level Actions:

• Issue warning

• Temporarily suspend account

• Permanently ban account

• Restrict specific features

Other Actions:

• Report to law enforcement (if illegal)

• Report to AI providers (for their records)

• Update filters to prevent similar content

9.2 Action Criteria

We consider:

• Severity of violation

• User's history

• Intent (malicious vs. accidental)

• Impact on others

• Legal requirements

9.3 Communication of Actions

You will be notified if:

• Your content is removed

• Your account receives a warning or suspension

• Action is taken on your report (sometimes)

Notification includes:

• What action was taken

• Reason for action

• How to appeal (if applicable)

9.4 Transparency in Enforcement

• We strive to explain our decisions

• Some actions cannot be disclosed (legal/privacy reasons)

• We publish general moderation statistics (when feasible)

10. TRANSPARENCY AND ACCOUNTABILITY

10.1 Transparency Report (Future)

We may publish periodic transparency reports including:

• Number of reports received

• Types of violations

• Actions taken

• Appeal outcomes

• System improvements

10.2 External Audit and Research

• We may engage external auditors to review moderation practices

• We support independent research on platform safety

• We are open to feedback and improvement

10.3 Public Engagement

• We welcome feedback on moderation policies

• We may consult users or experts on policy updates

• We are committed to ongoing improvement

10.4 Accountability to Regulators

• We comply with legal reporting requirements (e.g., DSA)

• We cooperate with law enforcement when legally required

• We respond to regulatory inquiries

11. LIMITATIONS OF MODERATION

⚠️ IMPORTANT – MODERATION IS NOT PERFECT ⚠️

11.1 We Cannot Catch Everything

Despite our efforts:

❌ Some harmful content may slip through

❌ Automated systems have false negatives

❌ Clever evasion attempts may succeed

❌ Human review cannot cover all content

❌ Context may be misunderstood

11.2 False Positives and False Negatives

False Positives:

• Safe content incorrectly flagged/removed

• We work to minimize these

• You can appeal

False Negatives:

• Harmful content incorrectly allowed

• We work to reduce these

• Report when you see them

11.3 No Guarantee of Safety

⚠️ We do NOT guarantee a completely safe or harm-free environment

⚠️ You may encounter content that disturbs or upsets you

⚠️ You use GioMind at your own risk

11.4 Technical Limitations

• AI is imperfect and evolving

• Moderation technology has limitations

• New types of harm emerge constantly

• We are always playing catch-up

11.5 Your Responsibility

• You are responsible for your own safety

• Use judgment when engaging with content

• Report violations when you see them

• Stop using features that cause distress

12. YOUR ROLE IN SAFETY

🤝 WE NEED YOUR HELP 🤝

12.1 Be a Responsible User

✓ Follow Community Guidelines

✓ Don't create harmful content

✓ Don't abuse AI features

✓ Respect others

12.2 Report Violations

✓ Report harmful content when you see it

✓ Report AI malfunctions or inappropriate responses

✓ Report bugs or security issues

12.3 Provide Feedback

✓ Tell us how we can improve

✓ Share ideas for better safety features

✓ Help us understand user needs

12.4 Take Care of Yourself

✓ Stop using GioMind if it causes distress

✓ Seek professional help if needed

✓ Don't rely on GioMind for safety-critical needs

13. APPEALS AND FEEDBACK

13.1 Appeal Process

If you disagree with a moderation decision:

• See Community Guidelines Section 12 for appeals process

• Email: giomind.app@gmail.com

• Subject: "Appeal – [Account Email]"

13.2 Feedback on Moderation

For general feedback:

📧 Email: giomind.app@gmail.com

Subject: "Moderation Feedback"

We review and consider all feedback.

14. CHANGES TO THIS POLICY

14.1 Updates

We may update this Content Moderation Policy to:

• Reflect new moderation technologies

• Address emerging threats

• Comply with legal requirements

• Improve clarity or effectiveness

14.2 Notice

• Updated policy will be posted in the app

• Material changes will be communicated via email or in-app notification

• Continued use constitutes acceptance

📋 SUMMARY – HOW WE MODERATE

Systems:

• Automated filters (pre and post AI)

• AI provider safety systems

• Human review (when needed)

• User reports

What We Moderate:

• Illegal content (zero tolerance)

• Harmful content (hate, violence, harassment)

• Dangerous misinformation

• Inappropriate AI use

• Platform abuse

Limitations:

• Cannot catch everything

• False positives and negatives occur

• No guarantee of complete safety

Your Role:

• Follow guidelines

• Report violations

• Provide feedback

Appeals:

• You can appeal moderation decisions

• Email giomind.app@gmail.com

🛡️ OUR COMMITMENT

We are committed to:

✓ Maintaining a safe platform

✓ Transparent moderation practices

✓ Continuous improvement

✓ Balancing safety and user freedom

✓ Accountability to users and regulators

Thank you for helping us keep GioMind safe and supportive.

📧 CONTACT

Questions or concerns about content moderation:

Email: giomind.app@gmail.com

Subject: "Content Moderation Question"

Last Updated: December 1, 2025