Content Moderation Policy
1. Overview
Quibit, Inc. ("Quibit," "we," "us," or "our") is committed to maintaining a safe, respectful, and lawful environment across all of our applications. This Content Moderation Policy describes how we detect, review, and act on content and behavior that violates our policies across all four Quibit, Inc. applications: Quibit (social), Kha (dating, 18+), QuiTalk (AI chat, channels, and phone calls), and QuizCoin (quiz game).
This policy applies to all forms of content and interaction on our platforms, including but not limited to: posts, comments, images, videos, messages, AI-generated content, AI conversations, dating profiles, phone calls, channel content, quiz submissions, in-game chat, live streams, and any other user-generated or system-generated content.
Our moderation practices are designed to be transparent, consistent, and fair while prioritizing user safety. We employ a combination of automated detection systems, AI-powered content analysis, and trained human reviewers to enforce our policies at scale.
Quibit enforces its content policies 24/7 across all applications. Violations are met with proportionate enforcement actions, ranging from content removal to permanent account termination and law enforcement referrals for severe offenses.
2. Moderation Framework
Quibit operates a multi-layered moderation framework that combines automated systems with human expertise to ensure comprehensive coverage across all applications and content types.
2.1 Core Principles
Our moderation framework is guided by the following principles:
- Enforcement actions are proportionate to the severity of the violation
- First-time minor violations receive warnings before escalation
- Severe violations such as CSAM, terrorism, or threats of violence result in immediate action
- The same rules apply to all users regardless of status, follower count, or subscription tier
- Moderation decisions follow documented guidelines and precedent
- Regular calibration sessions ensure reviewer alignment
- Users are notified when content is removed or accounts are restricted
- Reasons for enforcement actions are clearly communicated
- Moderation statistics are published in biannual transparency reports
- All users have the right to appeal moderation decisions
- Appeals are reviewed by a different moderator than the one who made the original decision
- Cultural and linguistic context is considered in moderation decisions
2.2 Moderation Layers
Our moderation system operates across three layers:
- Real-time content scanning using AI/ML classifiers
- Hash-matching for known violating content (PhotoDNA for CSAM)
- Keyword and pattern matching for spam, hate speech, and prohibited content
- Behavioral analysis for coordinated inauthentic behavior
- AI models classify flagged content by severity and violation type
- Automated prioritization ensures the most severe content is reviewed first
- AI provides moderators with context and recommended actions
- Continuous model improvement based on reviewer feedback
- Trained Trust & Safety specialists review escalated content
- Senior reviewers handle complex cases requiring nuanced judgment
- Subject matter experts are consulted for specialized content areas
- Quality assurance audits ensure accuracy and consistency
3. Content Review Process
All content on Quibit platforms is subject to review. Content may enter the review pipeline through automated detection, user reports, or proactive sweeps by our moderation team.
3.1 How Content Is Flagged
Content enters the moderation queue through the following channels:
Automated detection systems flag content that matches known violation patterns
User reports submitted via in-app reporting tools
Proactive sweeps by moderators targeting high-risk content areas
External reports from law enforcement, NGOs, or trusted flaggers
Cross-referencing with industry databases (e.g., NCMEC hash lists, terrorist content databases)
AI conversation monitoring detects harmful patterns in QuiTalk interactions
3.2 Review Prioritization
Flagged content is prioritized based on severity:
| Violation Type | Enforcement Action |
|---|---|
| CSAM / Child exploitation | Immediate review (within 1 hour). Content is hidden pending review. Reported to NCMEC upon confirmation. |
| Imminent threats of violence / Self-harm | Priority review (within 2 hours). Emergency services contacted if there is an immediate threat to life. |
| Terrorism / Violent extremism | Priority review (within 4 hours). Content removed and account suspended pending investigation. |
| Hate speech / Harassment / Bullying | Standard review (within 24 hours). Severe cases escalated to priority queue. |
| Spam / Scams / Misinformation | Standard review (within 24-48 hours). Automated systems may take interim action. |
| Community guideline violations | Standard review (within 48 hours). Low-severity content remains visible pending review. |
3.3 Review Outcomes
After review, content receives one of the following dispositions:
Approved — Content does not violate policies and is restored if previously hidden
Removed — Content violates policies and is permanently removed from the platform
Restricted — Content is age-gated, demoted in feeds, or limited in distribution
Labeled — Content receives an informational label (e.g., misinformation warning, sensitive content notice)
Escalated — Content requires additional review by senior moderators or subject matter experts
4. Automated Detection Systems
Quibit invests heavily in automated detection technology to identify policy-violating content at scale. Our systems process content in real time across all four applications.
4.1 Image and Video Analysis
Our visual content analysis systems include:
PhotoDNA hash-matching to detect known CSAM with near-zero false positive rates
Machine learning classifiers trained to detect nudity, violence, gore, and graphic content
Object detection models that identify weapons, drugs, and other prohibited items
Optical character recognition (OCR) to detect policy-violating text within images
Video frame analysis for real-time content in live streams and uploaded videos
Deepfake and synthetic media detection to identify manipulated content
4.2 Text and Language Analysis
Our text analysis systems cover all text-based content across platforms:
Natural language processing (NLP) classifiers for hate speech, harassment, and threats
Grooming and solicitation language detection for messaging and chat features
Spam and scam pattern recognition including phishing links and fraudulent schemes
Multilingual support covering English, Korean, Burmese, Thai, and Japanese
Context-aware analysis that considers conversation history and user relationships
Keyword and regex filters for prohibited terms, coded language, and evasion attempts
4.3 AI Conversation Monitoring (QuiTalk)
QuiTalk AI conversations receive specialized monitoring:
AI model outputs are filtered through safety classifiers before delivery to users
User prompts are screened for attempts to generate harmful, illegal, or exploitative content
Jailbreak and prompt injection attempts are detected and blocked in real time
AI-generated content is reviewed for accuracy, bias, and potential harm
Channel AI brain content undergoes automated safety review before activation
Conversation patterns indicating self-harm, radicalization, or exploitation trigger alerts for human review
AI phone call transcripts are monitored for abuse patterns and policy violations
See our Child Safety Policy for additional AI safety measures specific to minor protection.
4.4 Behavioral Analysis
Beyond content analysis, we monitor behavioral patterns to detect coordinated and systematic abuse:
Detection of coordinated inauthentic behavior (bot networks, fake engagement)
Account creation velocity monitoring to prevent spam account farms
Message frequency and pattern analysis to identify harassment campaigns
Network analysis to detect organized groups engaged in policy violations
Device fingerprinting to identify ban evasion and multi-accounting
Anomaly detection for unusual account activity (e.g., sudden spikes in messaging, mass following)
5. Human Review Team
Our Trust & Safety team consists of trained professionals who review content that requires human judgment. Automated systems handle the majority of clear-cut violations, while human reviewers focus on nuanced, context-dependent cases.
5.1 Qualifications and Training
All content moderators meet the following requirements:
Background checks and security clearance appropriate for handling sensitive content
Comprehensive initial training program (minimum 40 hours) covering all policy areas
Specialized training modules for CSAM identification, terrorism content, and crisis situations
Ongoing training on emerging threats, new evasion tactics, and policy updates
Language and cultural competency for markets served (Korean, Burmese, English, Thai, Japanese)
Annual recertification and performance evaluations
5.2 Moderator Wellbeing
We prioritize the mental health and wellbeing of our moderation team:
Access to licensed counselors and mental health professionals
Mandatory breaks and exposure limits for graphic content review
Rotation schedules to prevent prolonged exposure to harmful content
Peer support programs and team debriefing sessions
Optional reassignment to non-content-review roles
Comprehensive health benefits including mental health coverage
5.3 Escalation Procedures
Complex or high-severity cases follow a structured escalation path:
- Handle routine content reports and clear policy violations
- Apply standard enforcement actions per policy guidelines
- Escalate ambiguous or complex cases to Tier 2
- Review escalated cases requiring nuanced judgment
- Handle appeals of Tier 1 decisions
- Coordinate with legal team on borderline content
- Final decision authority on complex or high-profile cases
- Coordination with law enforcement and external agencies
- Policy interpretation and precedent-setting decisions
6. App-Specific Moderation
Each Quibit, Inc. application has unique content types and interaction patterns that require tailored moderation approaches. The following outlines app-specific moderation measures in addition to the universal policies described above.
6.2 Kha (Dating) — Profiles, Photos, and Interactions (18+)
Kha dating app content receives specialized moderation given its 18+ nature and interpersonal dynamics:
- Photo verification system requires users to submit real-time selfies matching their profile photos
- All profile photos are reviewed by automated systems and flagged for human review when necessary
- Profile text is scanned for prohibited content, solicitation, and underage indicators
- Bio content is checked for external contact information that may facilitate off-platform abuse
- Age verification is enforced at registration — users must be 18 or older
- Suspected underage users are immediately suspended pending identity verification
- Romance scam detection identifies patterns of financial exploitation and catfishing
- Unsolicited explicit photo detection automatically blurs and warns recipients
- Harassment and stalking patterns are detected through message analysis
- Location-based safety features include meeting point suggestions for public venues
- In-conversation reporting allows users to report matches without leaving the chat
- Post-date safety check-ins prompt users to report concerning behavior
- Blocked and reported users cannot re-match or view the reporter's profile
- Repeat offenders receive escalating restrictions up to permanent ban
See our Community Guidelines for behavioral expectations on Kha.
6.3 QuiTalk (AI Chat, Channels, and Phone Calls)
QuiTalk combines AI chat, user-created channels, and phone calling, each requiring distinct moderation approaches:
- AI model responses pass through multi-stage safety filters before delivery
- User prompts attempting to generate harmful content are blocked and logged
- Jailbreak attempts are detected, blocked, and tracked for pattern analysis
- AI conversations that indicate user distress trigger safety interventions and resource sharing
- AI-generated content is watermarked and labeled to prevent misuse
- Regular red-teaming exercises test AI safety boundaries and identify vulnerabilities
- Channel owners have access to moderation tools including message deletion, user muting, and banning
- Channel AI brain content is reviewed for compliance with content policies
- Channels promoting illegal activity, hate, or exploitation are removed and owners are actioned
- Public channel content is subject to the same automated scanning as Quibit social posts
- Channel FAQ and topic configurations are reviewed for policy compliance
- Channels with repeated violations may be permanently deleted with no restoration option
- Phone call metadata (duration, frequency, patterns) is analyzed for abuse indicators
- Users can report abusive calls through post-call reporting prompts
- Repeated harassment via phone calls results in call feature suspension
- Emergency call patterns (e.g., calls indicating distress) may trigger safety interventions
- Call recording and transcription, where legally permitted, may be used for safety investigations
6.4 QuizCoin (Quiz Game) — Content and Chat
QuizCoin game content and interactions are moderated to maintain a safe and fair gaming environment:
- User-submitted quiz questions are reviewed for accuracy, appropriateness, and policy compliance
- Quiz content containing hate speech, discrimination, or misinformation is rejected
- Automated screening prevents submission of quizzes with prohibited or harmful content
- Cultural sensitivity review ensures quiz content is appropriate across all markets
- Real-time chat filtering blocks profanity, slurs, and harassment
- Chat messages are monitored for grooming behavior and inappropriate contact with minors
- Spam and advertisement detection prevents disruption of game chat
- Users can mute, block, and report other players directly from the chat interface
- Cheating detection systems identify and action accounts using exploits or unauthorized tools
- Leaderboard manipulation and score tampering result in ranking removal and account suspension
- Coin farming and fraudulent reward claims are detected and reversed
7. Enforcement Actions
Quibit applies a graduated enforcement framework where the severity of the action corresponds to the severity and frequency of violations. The following table outlines standard enforcement actions, though Quibit reserves the right to take more severe action when warranted by the circumstances.
7.1 Graduated Enforcement Scale
For most policy violations, enforcement follows a graduated approach:
| Violation Type | Enforcement Action |
|---|---|
| First minor violation (e.g., mild incivility, borderline content) | Warning notification with education on community standards. Content may be removed. |
| Second minor violation within 90 days | Content removal and temporary restriction of posting privileges for 24-72 hours. |
| Third minor violation or first moderate violation (e.g., harassment, hate speech) | Content removal and account suspension for 7-30 days depending on severity. |
| Repeated moderate violations or first severe violation (e.g., threats, doxxing) | Extended account suspension (30-90 days) and permanent restriction of specific features. |
| Continued violations after suspension or extreme severity violation | Permanent account termination across all Quibit, Inc. applications. |
7.2 Immediate Action Violations
The following violations bypass the graduated enforcement scale and result in immediate action:
| Violation Type | Enforcement Action |
|---|---|
| CSAM or child exploitation content | Immediate permanent ban across all apps. Content preserved and reported to NCMEC and law enforcement within 24 hours. |
| Credible threats of violence or terrorism | Immediate account suspension. Content removed and law enforcement notified. Permanent ban upon confirmation. |
| Distribution of non-consensual intimate images | Immediate content removal and account suspension. Permanent ban for repeat offenders. |
| Human trafficking or exploitation | Immediate permanent ban. Evidence preserved and reported to law enforcement and relevant authorities. |
| Sale of illegal drugs, weapons, or controlled substances | Immediate content removal and permanent account ban. Reported to law enforcement. |
| Impersonation of law enforcement or government officials | Immediate content removal and account suspension pending investigation. |
See our Child Safety Policy for detailed CSAE enforcement procedures.
7.3 Feature-Specific Restrictions
In addition to account-level actions, Quibit may restrict access to specific features:
Posting restrictions — User cannot create new posts, comments, or stories
Messaging restrictions — User cannot send direct messages or participate in group chats
Channel restrictions — User cannot create new channels or post in existing channels
Phone call restrictions — User cannot make or receive calls through QuiTalk
AI chat restrictions — User's AI usage is suspended or limited
Matching restrictions — User cannot match with or message new users on Kha
Game restrictions — User is excluded from competitive play and leaderboards on QuizCoin
Discovery restrictions — User's profile and content are hidden from search and recommendations
8. Appeals Process
Quibit believes in fair and transparent enforcement. All users whose content is removed or whose accounts are restricted have the right to appeal, except in cases involving CSAM or confirmed child exploitation.
8.1 How to Submit an Appeal
Users can appeal moderation decisions through the following methods:
In-app appeal button — Available on all enforcement notifications within the app
Email appeal — Send a detailed appeal to [email protected] with your account ID and the decision reference number
Web form — Submit an appeal through the Appeals Center at thequibit.com/appeals
Appeals must be submitted within 30 days of the enforcement action. Appeals submitted after 30 days may not be reviewed.
8.2 Appeal Requirements
To ensure timely processing, appeals should include:
Account identifier (username or registered email address)
Reference number of the enforcement action (included in the notification)
Explanation of why the user believes the decision was incorrect
Any relevant context or evidence supporting the appeal
Confirmation that the user has read and understands the applicable policy
8.3 Review Process and Timelines
Appeals are reviewed by a different moderator than the one who made the original decision:
- Reviewed within 5 business days of submission
- User is notified of the outcome via email and in-app notification
- Decisions may be upheld, modified, or overturned
- Cases requiring additional investigation are reviewed within 15 business days
- User is notified of the extended timeline and given an updated estimate
- May involve consultation with legal counsel or subject matter experts
- Users who disagree with the appeal outcome may request a final review by Trust & Safety leadership
- Final appeals are reviewed within 20 business days
- The final appeal decision is binding and concludes the review process
8.4 Non-Appealable Decisions
The following enforcement actions are not eligible for appeal:
Accounts terminated for confirmed CSAM or child exploitation
Accounts terminated pursuant to law enforcement requests or court orders
Accounts terminated for confirmed terrorism or violent extremism
Temporary content removals during active investigations (content may be restored upon completion)
9. Transparency Reporting
Quibit is committed to transparency about our moderation practices and their impact. We publish regular reports to inform users, regulators, and the public about our enforcement activities.
9.1 Report Contents
Our biannual transparency reports include the following metrics:
Total volume of content reviewed (automated and human review)
Breakdown of violations by category (hate speech, harassment, spam, CSAM, etc.)
Enforcement actions taken by type (warnings, suspensions, terminations)
Proactive detection rate vs. user-reported content ratio
Average review time by severity category
Appeal volume, outcomes, and overturn rates
Government and law enforcement requests received and fulfilled
NCMEC CyberTipline reports submitted
App-specific moderation statistics for each Quibit, Inc. application
9.2 Publication Schedule
Transparency reports are published on the following schedule:
Biannual reports — Published in January and July covering the preceding 6-month period
Annual summary — Published in January with year-over-year comparisons and trend analysis
Incident reports — Published as needed following significant safety incidents or policy changes
Reports are available at thequibit.com/transparency and archived for public access
9.3 Independent Audits
Our moderation practices are subject to external review:
Annual third-party audits of moderation accuracy and consistency
Independent review of AI/ML model fairness and bias
Regulatory compliance assessments as required by applicable laws
Engagement with academic researchers on content moderation effectiveness
10. User Reporting Tools
Users are essential partners in maintaining safe communities. Quibit provides accessible reporting tools across all applications to empower users to flag content and behavior that may violate our policies.
10.1 Reporting Tools by Application
Each application provides contextual reporting options:
- Report button on every post, comment, story, and profile
- Long-press on messages to report individual messages in conversations
- Report options include: spam, harassment, hate speech, violence, misinformation, nudity, and other
- Anonymous reporting — reported users cannot see who filed the report
- Report and block available on every profile and in every conversation
- Post-date safety check-in prompts with easy reporting access
- Specific report categories for romance scams, catfishing, underage users, and harassment
- Emergency report option for immediate threats to personal safety
- Report button on AI-generated responses that are harmful or inaccurate
- Channel message reporting for inappropriate content from other users
- Post-call reporting for abusive phone calls
- Channel reporting for channels promoting prohibited content
- AI safety feedback allows users to flag AI responses for review and model improvement
- Report button on quiz content for inaccurate, offensive, or inappropriate questions
- In-game chat reporting for harassment, spam, and inappropriate messages
- Player profile reporting for offensive usernames, avatars, or bios
- Fair play reporting for suspected cheating or exploitation
10.2 Report Handling
All user reports are handled according to the following process:
Reports are acknowledged automatically upon submission
Each report is assigned a unique reference number for tracking
Reports are prioritized based on severity and reviewed according to the timelines in Section 3
Reporters receive a follow-up notification when their report has been reviewed
Repeated false or malicious reports may result in restrictions on the reporting user's account
Users who report in good faith are never penalized, even if the report does not result in enforcement action.
10.3 Trusted Flagger Program
Quibit partners with trusted organizations and community leaders to enhance report quality:
NGOs, academic institutions, and safety organizations can apply for trusted flagger status
Trusted flagger reports receive priority review and expedited processing
Regular feedback loops ensure trusted flaggers are calibrated with current policies
Government agencies and law enforcement may submit reports through dedicated channels
11. Cross-App Enforcement
Quibit, Inc. operates a unified trust and safety infrastructure across all four applications. Policy violations on one app may result in enforcement actions across all apps, ensuring that bad actors cannot simply move to another Quibit application to continue violating policies.
11.2 Cross-App Enforcement Scenarios
The following scenarios illustrate how cross-app enforcement operates:
| Violation Type | Enforcement Action |
|---|---|
| User permanently banned from Kha for harassment | Account is suspended on Quibit, QuiTalk, and QuizCoin pending review. Permanent ban applied across all apps if violation history warrants it. |
| User caught distributing CSAM on Quibit | Immediate permanent ban across all four applications. All content removed. Reported to NCMEC and law enforcement. |
| User receives multiple warnings on QuizCoin for hate speech | Warning record is visible to moderators across all apps. Escalation thresholds account for violations on any app. |
| User creates new account on QuiTalk after being banned from Quibit | Ban evasion detected through device fingerprinting. New account is immediately terminated. |
11.3 Cross-App Data Sharing for Safety
To enable effective cross-app enforcement, the following data is shared across applications:
Violation history and enforcement actions
Device identifiers and fingerprints
Content hashes of removed violating content
Behavioral risk scores and account flags
Block and report records
Cross-app data sharing for safety purposes is described in our Privacy Policy. User data is never shared with third parties for advertising purposes.
12. Contact and Policy Updates
For questions, concerns, or feedback about this Content Moderation Policy or our moderation practices, please contact us through any of the following channels.
12.1 Policy Update Process
This Content Moderation Policy is reviewed quarterly and may be updated to reflect:
Changes in legal or regulatory requirements across jurisdictions
New content types or platform features requiring moderation
Improvements to detection technology and moderation practices
Feedback from users, civil society organizations, and regulators
Evolving threat landscape and emerging abuse patterns
Industry best practices and standards
Material changes to this policy will be communicated via email and in-app notifications at least 30 days before taking effect. Continued use of any Quibit, Inc. application after the effective date constitutes acceptance of the updated policy.
For questions about this policy, contact [email protected]
© 2026 Quibit, Inc. · Version 1.0· Last updated 2026-03-13