​
What is the Omnifact Privacy Filter™?

Our Omnifact Privacy Filter™ is an AI safety tool that automatically detects and anonymizes sensitive information in your messages and documents before they’re processed by AI models. It’s especially valuable when using non-EU or third-party models where data privacy is a concern.

The Privacy Filter is enabled by default across the Omnifact platform and serves as an additional layer of protection against accidental data sharing.

​
The Privacy Filter Process

1

Finding Sensitive Information

When you send a message or upload documents, the Privacy Filter immediately checks for sensitive information like names, phone numbers, and addresses.

This check happens within Omnifact’s secure systems before any information reaches external AI models.

2

Filtering the Sensitive Parts

The Privacy Filter replaces sensitive information with generic, unidentifiable labels while preserving the meaning of your message.

These labels keep your message readable while protecting the private details from external AI models.

3

Safe AI Processing

Your message is processed using only the filtered version, ensuring your private information never reaches external models.

4

Automatic Re-insertion

When you receive a response, the Privacy Filter automatically restores your original information in place of the labels. This happens seamlessly, making it appear as if no filtering occurred.

If the identity of filtered terms is essential for a meaningful answer (e.g., asking a specific question about a person), the AI model won’t have access to this information and may not be able to provide a helpful response. In such cases, you may need to selectively reveal certain information using Click to Reveal.

​
What You’ll See When Information Gets Filtered

When you send a message with sensitive information, you’ll notice some changes in how your message appears:

In your message: You’ll see certain words or phrases highlighted with a special visual indicator (like a crossed-out eye icon) where your sensitive information was detected. This shows you exactly what information was protected.

What this means: If you see these highlighted areas in your message, it means the Privacy Filter detected sensitive information and is protecting it from being shared with the AI assistant.

In the screenshot above: You can see how a message that originally contained specific names and company information now shows highlighted areas with mask icons. The person’s name and company name are visually marked as protected, while the rest of the message remains unchanged.

While the Privacy Filter provides protection, it does not replace responsible AI usage. Users should still exercise caution with the information they share.

​
Types of Sensitive Information Detected

The Privacy Filter is designed to detect and a wide range of sensitive information:

  • Full names
  • Individual names
  • Email addresses
  • IP addresses
  • Phone numbers
  • Personal addresses
  • Social Security Numbers
  • Driver’s license numbers
  • ID card/passport numbers

The Privacy Filter is regularly updated to recognize new types of sensitive information.

​
Responsible AI Usage

While the Privacy Filter provides technical protection, Omnifact Academy helps users build confidence in working with AI through comprehensive training. Users learn how to effectively leverage AI capabilities while maintaining security best practices, making adoption smoother across your organization. This combination of technical safeguards and practical knowledge empowers teams to embrace AI tools with confidence.

​
Next Steps

Now that you understand how the Privacy Filter works, you might want to explore: