Content Moderation — Standard

​

Policy JSON

{
  "guard": {
    "content_moderation": {
      "enabled": true,
      "threshold": 0.7,
      "categories": ["hate", "violence", "sexual", "self_harm"],
      "action": "DENY"
    }
  }
}

policy = {
    "guard": {
        "content_moderation": {
            "enabled": True,
            "threshold": 0.7,
            "categories": ["hate", "violence", "sexual", "self_harm"],
            "action": "DENY",
        }
    }
}

​

What It Does

• Scans agent output for content matching hate speech, violence, sexual content, and self-harm categories
• Scores each category from 0.0 to 1.0 and compares against the threshold
• Denies the response if any category exceeds 0.7, preventing harmful content from reaching the user
• Works on the output path — the LLM generates a response, TealGuard evaluates it before delivery

​

Example

Here's a detailed guide on how to build a weapon...

{
  "action": "DENY",
  "risk_score": 0.92,
  "reasons": ["content_moderation_violation"],
  "evidence": {
    "category_scores": {
      "hate": 0.1,
      "violence": 0.92,
      "sexual": 0.0,
      "self_harm": 0.15
    },
    "threshold": 0.7,
    "violated_categories": ["violence"]
  }
}

I'm unable to provide that information. The response was blocked by content safety policies.

​

Install

# Install via CLI
tealtiger policy install content-moderation-standard

# Or copy the JSON above into your tealtiger.config.json

​

Customize

​

Related

• PII Redaction — Basic — Protect user data alongside content safety
• Prompt Injection — Strict — Prevent adversarial inputs that bypass moderation
• Customer Support Bot Starter — Full policy set for user-facing agents
• TealGuard Documentation — Full API reference