Moderation teams can only scale so far. The Online Safety Toolkit gives Verint Community a layered, automated safety system that catches harmful content, malicious links, and network-level threats before they reach your members or your moderators.

Content safety

AI content Moderation

The toolkit's text analysis engine uses a custom-trained AI model to read posts and private messages and detect bullying, harassment, and toxic language. This is not a keyword blocklist. The model understands context, so it catches subtle abuse that trigger-word filters miss, and avoids flagging legitimate discussion that happens to contain sensitive language.

When the model flags content, it passes a verdict to the moderation engine, which can hide the content, notify a moderator, or both. Your team reviews edge cases; the obvious stuff never reaches them.

Image and video Moderation

Three moderation plugins cover visual content across different cloud providers, so you can use whichever your organisation already has a relationship with.

• AWS Image Moderation — scans uploaded images via AWS Rekognition for explicit nudity, violence, and offensive symbols. Confidence thresholds are configurable per category so you control how strict each one is.
• AWS Video Moderation — extends image moderation to video uploads. Files are processed asynchronously via S3 with configurable polling and retry logic, so the community stays responsive while scans run.
• Google Image Moderation — uses Google Cloud Vision to classify images across adult, medical, violence, spoof, and racy categories. Works as an alternative to AWS or alongside it.

Both AWS and Google plugins use your own cloud credentials. Images and videos flow from your community to your own cloud account, not through 4 Roads.

Link safety

UrlHaus link checker

Every link posted by a user is checked against the UrlHaus malware distribution database before the content is published. The local threat list updates automatically on a schedule, so protection stays current without manual intervention. Members cannot accidentally share links that download viruses or ransomware.

Link safety modifier

A simple but effective layer of protection for members clicking outbound links. The plugin can:

• Force all external links to open in a new tab, preventing unexpected page navigation
• Route external links through a "you are leaving this site" interstitial page, so members know before they go

Both options are independently configurable.

Network security

Malicious IP blocker

Before a user can post content, their IP address is checked against the Ipsum threat feed of known spam and malware sources. If the IP is on the list, the content is flagged automatically. The plugin supports trusted proxy CIDR ranges so resolution is accurate behind load balancers and reverse proxies.

HTTP request logger

Every HTTP request to your community is logged: method, IP address, path, user agent, referrer, status code, and timestamp. Logs are stored securely and are downloadable by administrators. When a security incident needs investigating, you have the full audit trail to reconstruct what happened and when.

Observability

OST audit data can ship directly to your existing monitoring stack. Three adapters are included:

• DataDog — configure your API key, region endpoint, and service name
• Elasticsearch — configure your endpoint, index name, and API key
• Grafana Loki — configure your endpoint, user ID, and auth token

Log level and output are controlled through a central logging configuration plugin.

Your data stays private

The AI content labeling engine is powered by the Contextuly API, which runs on infrastructure owned and operated by 4 Roads. Community content is never sent to consumer AI services such as OpenAI, Google AI, or Anthropic.

• Content submitted for labeling is processed on 4 Roads' own hosted servers
• Nothing is stored after analysis completes
• Your data is never used to train or improve any model
• The AWS and Google image moderation plugins use your own cloud provider credentials — data flows from your community to your own cloud account only

All AI-powered plugins require explicit administrator acknowledgement before activation, confirming you understand what data is processed and where.

What it means for your team

For community managers: harmful content is flagged or removed automatically before it reaches the review queue. You spend time building community, not policing it.

For platform administrators: every threat vector has a response. Malicious IPs are blocked before they post. Phishing links are caught before members click them. Every request is logged for incident investigation. When something goes wrong, you have the evidence.

For the business: a defensible safety posture, a reduced liability exposure from user-generated content, and a foundation for demonstrating compliance with online safety obligations.

Requirements

• Verint Community 12 or later
• AI Content Labeling requires a Contextuly API subscription
• AWS Image/Video Moderation requires an active AWS account with Rekognition access
• Google Image Moderation requires a Google Cloud account with Vision API enabled