Creating comprehensive moderation policies from scratch can be time-consuming and complex. DeepMod's automatic policy generation feature streamlines this process by transforming your existing policy documents into structured, ready-to-use moderation policies powered by advanced AI.

What Is Automatic Policy Generation?

Automatic policy generation allows you to upload an existing policy document (such as community guidelines, content standards, or moderation rules) and have DeepMod automatically extract, organize, and structure it into policies, rule groups, and individual rules that are ready to use in your moderation workflows.

Instead of manually creating policies rule by rule, you can leverage documents you've already written to jumpstart your moderation setup.

Don't have existing policy documents? Consider browsing policy templates instead (navigate to Policies → Templates). DeepMod offers pre-built, industry-specific policies that you can clone instantly, offering comprehensive coverage without requiring source documents.

How It Works

The policy generation process uses advanced language models to understand your document's content and intelligently organize it into DeepMod's hierarchical policy structure.

The Generation Process

1. Document Upload

Upload your policy document as a PDF file (up to 10MB). The system accepts various document types including community guidelines, terms of service, content policies, brand guidelines, and moderation standards.

2. Text Extraction & Analysis

DeepMod extracts and sanitizes the text from your document, removing formatting artifacts while preserving the meaning and structure of your policy content.

3. Intelligent Structuring

Our AI analyzes your document in three stages:

• Rule Extraction: Individual, actionable rules are identified from your content

• Group Organization: Rules are organized into logical groups (Safety, Legal, Brand, etc.)

• Policy Creation: Groups are assembled into coherent policies by theme or purpose

4. Policy Creation

The system creates your policies in DeepMod with:

• Clear, descriptive names and descriptions

• Properly organized rule groups

• Well-defined individual rules

• Your specified confidence threshold and review mode settings

All generated policies start in an inactive state, allowing you to review, test, and refine them before activating for production use.

Generation Limits

Automatic policy generation has the following constraints:

Constraint	Limit
Maximum policies per generation	5
Maximum rule groups per policy	50
Maximum rules per policy (Free tier)	25
Maximum rules per policy (Paid tier)	100
Maximum file size	10MB
Supported format	PDF

Field length limits:

• Policy names: 2-100 characters

• Policy descriptions: 10-500 characters

• Rule group names: 2-160 characters

• Rule group descriptions: 10-500 characters

• Rule names: 5-80 characters

• Rule content: 5-160 characters

Content exceeding these limits will be trimmed or split during generation.

Using Automatic Generation

Step 1: Prepare Your Document

For best results, ensure your policy document:

• Is well-organized with clear sections and topics

• Contains specific rules rather than just general principles

• Uses clear language about what content should or shouldn't include

• Is in PDF format for upload

Step 2: Navigate to Policy Generation

In your DeepMod dashboard:

1. Navigate to the Policies section

2. Click the Generate from Document button

3. This opens the policy generation interface

Step 3: Upload and Configure

Upload Your Document:

• Drag and drop your PDF file or click to browse

• Maximum file size: 10MB

• Supported format: PDF

Set Confidence Threshold:

• Use the slider to choose how certain the AI must be before making definitive decisions (1-100%)

• Start with 70-80% for a balanced approach

• You can adjust this later for individual policies after generation

Enable Human Review (Optional):

• Toggle this on if you want ambiguous cases flagged for human review

• Leave it off for fully automated moderation

• This setting applies to all generated policies but can be changed individually later

Step 4: Submit for Generation

Click Generate Policies to begin the process. You'll receive immediate confirmation that generation has started, and you'll be redirected to the policies page where you can monitor progress.

Step 5: Monitor Progress

Policy generation happens asynchronously in the background. A status card appears on the Policies page showing real-time progress through these stages:

Status	Description
Queued	Generation request received, waiting to process
Processing Rules	AI is extracting individual rules from your document
Processing Groups	AI is organizing rules into logical groups
Processing Policies	AI is assembling groups into final policies
Completed	Generation finished successfully
Failed	An error occurred during generation

The status card updates automatically and shows counts of generated policies, groups, and rules as they're created.

Note: If a generation remains in "Queued" status for more than 15 minutes, it will automatically be marked as failed. You can delete failed generations and try again.

Understanding Generated Policies

What You'll Get

After successful generation, you'll find new policies in your dashboard (up to 5 per generation), each containing:

Policy-Level Information:

• Descriptive name reflecting the policy's focus area

• Detailed description explaining the policy's purpose

• Your specified confidence threshold and review mode

• Unique URI for API access

• Inactive status (ready for your review)

Organized Rule Groups:

• Logical groupings like Safety, Legal, Brand, Spam, etc.

• Clear descriptions of each group's focus

• Multiple rules within each group

Specific Rules:

• Actionable, testable rules extracted from your document

• Clear rule names and detailed content descriptions

• Language optimized for AI understanding

After Generation: Next Steps

Generated policies require review and activation before they can moderate content. Follow these steps to put your policies into production:

1. Review Generated Content

Navigate to each generated policy and examine:

• Overall policy organization and structure

• Rule group categorization

• Individual rule clarity and accuracy

• Any redundant or overlapping rules

2. Refine as Needed

While AI does an excellent job extracting and organizing rules, you may want to:

• Edit rule wording for clarity

• Merge similar or redundant rules

• Add rules for edge cases not covered in your document

• Adjust rule group organization

• Delete any irrelevant or incorrectly extracted rules

3. Test Thoroughly

Before activating, test each policy with representative content:

• Use the policy testing interface

• Submit 10-20 diverse content samples

• Review results for accuracy and appropriateness

• Adjust confidence thresholds if needed

• Refine rule wording based on test outcomes

4. Activate Policies

Once satisfied with a policy's performance:

• Change its status from Inactive to Active

• The policy becomes available for moderation requests

• Begin monitoring real-world performance

5. Monitor and Iterate

After activation:

• Review moderation results regularly

• Check for unexpected outcomes or patterns

• Refine rules based on real-world usage

• Adjust confidence thresholds as needed

Managing Generations

Viewing Generation History

Active generations (not yet completed) appear as status cards on the Policies page. Once completed, the generated policies appear in your standard policy list.

Deleting Failed Generations

If a generation fails, you can delete it by clicking the delete button on the status card. Only failed generations can be deleted—completed generations result in actual policies that should be managed through the normal policy interface.

Best Practices

Start with quality source material:

• Well-organized documents produce better results

• Clear section headers help the AI understand structure

• Specific language about content requirements translates better than vague principles

Focus your documents:

• Single-purpose documents work better than combining multiple policy areas

• If you have separate documents for different policy types (safety, legal, brand), generate from each separately

• This produces more focused, maintainable policies

Limitations and Considerations

What Works Well

Automatic generation excels at:

• Extracting explicit rules and standards

• Organizing content into logical groups

• Converting existing policy documents into structured formats

• Providing a strong starting point for new moderation policies

What Requires Manual Work

Some aspects still require human judgment:

• Nuance and context: AI may not capture subtle policy intentions

• Edge cases: Unusual situations may need manual rule additions

• Optimization: Fine-tuning thresholds and rule wording for your specific content

• Validation: Ensuring extracted rules accurately reflect your requirements

Privacy and Security

• Files are not stored: After text extraction, uploaded files are discarded

• Text is temporary: Extracted text exists only during the generation process

• Policies are private: Generated policies belong to your organization only

What's Next

After successfully generating policies from your documents, explore these topics:

Learn how to refine and optimize generated policies

• Policy Authoring Best Practices - Learn how to refine and optimize generated policies

• Running Moderation Effectively - Put your policies into production

Automatic policy generation gives you a powerful head start on content moderation. Combined with testing, refinement, and ongoing iteration, it enables you to deploy sophisticated moderation policies quickly and efficiently.

Questions about automatic policy generation? Contact our support team for assistance with document preparation, generation issues, or policy refinement strategies.