Content Moderation Framework

1. Purpose & Scope

Vibble is a global microblogging and conversational platform where users can post short text updates, media, and links in real time. The speed, reach, and public nature of content on Vibble create unique safety, integrity, and moderation challenges. This framework explains:

• How content and accounts are reviewed and moderated
• Which signals and tools are used (automated and human)
• How enforcement decisions are made, documented, and appealed
• How Vibble mitigates systemic risks under applicable laws and regulations

This framework applies to:

• All posts (“vibes”), replies, quotes, reposts, threads, DMs (where applicable), media uploads, and profile elements
• All users and organizations, including verified accounts, brands, media, and government entities
• All enforcement activity performed by or on behalf of Vibble and Nexa Group

2. Moderation Objectives

Vibble’s moderation ecosystem is driven by four core objectives:

1. Safety: Prevent and mitigate harm, including harassment, threats, exploitation, and child endangerment.
2. Integrity: Reduce manipulation, spam, and coordinated influence operations that distort authentic conversation.
3. Compliance: Meet or exceed legal and regulatory obligations across all operating jurisdictions.
4. Rights & Fairness: Respect freedom of expression, due process, and non-discrimination in enforcement.

3. Policy Foundations

Enforcement decisions under this framework are grounded in the following policy sets:

• Community Guidelines (including Harassment & Abuse, Hate Speech, Threats & Extremism)
• Safety, Child Protection & CSAM Policies
• Election Integrity & Political Content Policy
• Sensitive Media, Adult Content, and Graphic Imagery Policies
• Spam, Bot, and Platform Manipulation Policies
• Identity, Verification & Account Authenticity Policies
• Legal & Regulatory Policies (including DSA, Online Safety Act, GDPR)

This framework does not replace those policies; it operationalizes how they are implemented.

4. Moderation Architecture: Multi-Layer Model

Vibble uses a multi-layer moderation model designed specifically for real-time, high-volume posting:

1. Pre-Publication & Inline Signals – Basic filters and heuristics applied at the moment of posting (e.g., CSAM hash-matching, known banned terms, URL blocklists).
2. Post-Publication Automated Review – Machine learning (ML) models score public posts, replies, quotes, and media for risk.
3. User & Partner Reports – Flagging from users, trusted reporters, NGOs, and verified partners feeds dedicated review queues.
4. Human Review & Specialist Escalation – Safety specialists and policy experts adjudicate complex, high-risk, and appellate cases.
5. Systemic Monitoring & Risk Assessments – Ongoing analysis for coordinated abuse, election manipulation, and cross-platform campaigns.

5. Automated Detection & Scoring

Vibble deploys multiple ML and rules-based systems to detect potential violations:

• Text Classification: Harassment, hate speech, threats, self-harm content, spam, and political misinformation cues.
• Media Analysis: Detection of nudity, sexual content, graphic violence, and known CSAM hashes.
• Behavioral Models: Spam, bot behavior, coordinated posting, reply-hijacking, and engagement manipulation.
• Network & Graph Analysis: Identification of inauthentic clusters, brigading, and abuse rings.

Each content item may receive multiple scores:

• Safety Risk Score (e.g., harassment, self-harm, child risk)
• Integrity Risk Score (e.g., spam, automation, manipulation)
• Sensitivity Score (e.g., graphic media, adult imagery)

High-risk scores can trigger:

• Immediate hidden or limited visibility pending review
• Auto-removal for certain clearly illegal or zero-tolerance categories (e.g., CSAM)
• Escalation to human review queues with priority tags

6. Human Moderation & Specialist Teams

Human reviewers are essential for nuanced, context-sensitive decisions. Vibble’s moderation organization includes:

• Core Moderation Team: Handles the majority of daily violation reviews, guided by detailed enforcement playbooks.
• Child Safety & CSAM Specialists: Trained for extremely sensitive and legally regulated content.
• Harassment & Targeted Abuse Experts: Focused on complex harassment, dogpiling, brigading, and doxxing.
• Election & Civic Integrity Specialists: Oversee political and election-related content, including misinformation and state-linked operations.
• Legal & Escalations Team: Coordinates with Nexa Group legal counsel, regulators, and law enforcement.
• Regional & Language Moderators: Provide cultural and linguistic context for content in non-English languages and local political environments.

7. Triage, Queues & Priority Levels

All flagged content (automated or reported) is routed into queues with priority based on:

• Severity (e.g., child safety, threats vs. mild harassment)
• Reach (e.g., trending posts, verified or government accounts)
• Context (e.g., ongoing crisis, election period, coordinated campaign)

Vibble uses an internal triage matrix that defines:

• P0 – Critical: Imminent harm, CSAM, terrorism, explicit violent threats; near-real-time handling.
• P1 – High: Severe harassment, doxxing, hate speech, large-scale manipulation.
• P2 – Medium: Repeated abuse, misinformation with potential harm, spam networks.
• P3 – Low: Isolated minor policy violations, borderline content, context clarification.

8. Enforcement Actions Spectrum

Vibble applies proportional enforcement based on severity, user history, and risk:

• Content-Level Actions
  • Labeling (e.g., “potentially sensitive”, “context added”, “misleading”)
  • Demotion or reduced visibility in feeds, search, and trends
  • Limiting ability to repost, reply, or quote
  • Removal of individual posts, media, or threads
• Account-Level Actions
  • Warnings and education notices
  • Reply, repost, or messaging restrictions
  • Temporary read-only mode
  • Temporary suspension
  • Permanent account suspension and content removal
• Systems-Level Actions
  • Shadow-limits for spam and manipulation patterns
  • Graph-wide demotion of coordinated networks
  • Blocking of IPs, device fingerprints, and automation endpoints for abusive actors

9. Shadowbanning & Visibility Reduction

To protect the platform from spam, brigading, and malicious amplification, Vibble may reduce the visibility of specific posts or accounts without explicitly notifying the actor when:

• Automated systems detect high confidence of spam or coordinated inauthentic behavior
• Content is borderline harmful or misleading but does not meet removal thresholds
• Accounts repeatedly post toxic, abusive, or low-quality content across multiple reports

Visibility reduction is governed by the separate Shadowban & Visibility Reduction Policy, which defines thresholds, durations, and appealable conditions.

10. Special Handling: Verified, Government & High-Reach Accounts

Posts from verified individuals, organizations, and government entities can have outsized impact. For such accounts:

• Certain enforcement actions (especially removals) undergo additional review layers.
• Public interest and newsworthiness are weighed carefully against potential harm.
• Public interest labels, context notices, or limited interactivity may be used instead of removal, where appropriate and safe.
• During elections or crises, stricter civic integrity standards may apply.

11. Misinformation & Civic Integrity

Vibble’s moderation of misinformation focuses primarily on content that can cause tangible harm:

• Misleading information about voting procedures and election integrity
• Dangerous public health misinformation during crises
• False narratives that facilitate fraud, scams, or organized harm

Possible actions include:

• Context labels and links to authoritative sources
• Reduced amplification or removal from trends and recommendations
• Account restrictions for persistent, high-impact misinformation campaigns

12. Harassment, Abuse & Doxxing

For harassment and targeted abuse:

• Vibble considers patterns across multiple posts, replies, and quote-posts.
• Dogpiling (mass participation in targeted harassment) is treated as coordinated abuse.
• Doxxing (sharing private personal information) is subject to expedited removal and heightened penalties.
• High-risk targets (minors, vulnerable groups, public figures facing threats) are prioritized in review.

13. Child Safety & CSAM

Child safety is treated as a zero-tolerance domain:

• CSAM and child sexual exploitation triggers immediate removal, permanent bans, and reporting to appropriate authorities.
• Systems use hash-matching, heuristics, and human review to detect and prevent re-uploads.
• Accounts suspected of grooming behavior or predatory conduct are escalated to specialist teams.

14. Bot, Automation & Spam Controls

Vibble applies layered defenses against spam and automation:

• Rate limits on posting, following, messaging, and engagement for new or suspect accounts.
• Device and IP fingerprinting to identify abusive scripts and automation.
• Graph analysis to detect engagement pods, link farms, and coordinated promotion networks.
• Shadow-limits or removal of content from known spam clusters.

15. Moderation of Replies, Quotes & Trends

Because much abuse occurs in replies and quote-posts, Vibble:

• Monitors reply patterns for brigading and harassment.
• Allows creators and users to restrict who can reply or quote their posts.
• Applies trending and search safeguards to prevent harmful or low-quality content from dominating discovery surfaces.

16. Documentation, Logging & Audit Trails

For every significant moderation decision, Vibble maintains structured logs including:

• Content identifiers and timestamps
• Automated model scores and rules triggered
• Moderator decision, rationale, and policy references
• Escalation chain, if any
• Appeal submissions and outcomes

These logs support:

• Internal quality review and calibration
• Regulatory audits and transparency reporting
• Forensic analysis and law enforcement cooperation where legally required

17. Appeals & Error Correction

Users and organizations may contest decisions via the Vibble Appeals System. Appeals:

• Are reviewed by a moderator who did not make the original decision.
• May trigger additional fact-finding or policy clarification.
• Can result in reversal, partial modification, or affirmation of the decision.

Insights from successful appeals inform:

• Policy updates and clarifications
• Model retraining and threshold tuning
• Moderator training and calibration exercises

18. Quality Assurance, Training & Calibration

Vibble runs continuous quality programs to ensure moderation consistency:

• Regular sampling and review of moderator decisions
• Calibration sessions across regions and teams
• Training modules on policy updates, emerging harms, and regional developments
• Bias and non-discrimination training for human reviewers

19. Risk Assessments & Regulatory Compliance

Under applicable laws (including the EU Digital Services Act and UK Online Safety Act), Vibble:

• Conducts regular systemic risk assessments on child safety, disinformation, illegal content, and civic discourse.
• Documents mitigation measures implemented within recommendation systems, trending, and search.
• Produces transparency and enforcement reports with aggregated statistics.
• Engages with regulators, civil society, and external experts where appropriate.

20. Continuous Improvement & Roadmap

Moderation systems evolve with user behavior and global events. Vibble’s roadmap includes:

• Expanding language and regional coverage for moderation
• Enhancing explainability of automated decisions to users
• Tighter integration of user feedback into product and policy design
• Ongoing investment in Trust & Safety engineering, operations, and governance at Nexa Group level

21. Contact & Escalation Channels

For questions about this framework, moderation practices, or escalations:

Vibble Safety Team: safety@vibble.com
Policy & Enforcement: policy@vibble.com
Appeals: appeals@vibble.com
Nexa Group Legal: legal@nexa-group.org

22. Updates to This Framework

Vibble and Nexa Group may update this Content Moderation Framework to reflect new regulations, product changes, emerging risk categories, or operational learnings. When material changes occur, we will:

• Update this page with a new “Last Updated” date
• Reflect changes in related policy documents (e.g., Community Guidelines, Safety Center)
• Communicate material updates to users or partners where legally required or operationally appropriate