Handling Different Content Types

In Clavata.ai, Policies govern the content on your platform. This means you don’t need to create distinct Policies for each type of content since one Policy can address both.

Clavata.ai eliminates the need for lengthy keyword lists or known problematic hashes by allowing you to define adaptable Policies for AI-driven content analysis. This makes the system more flexible and less dependent on predefined terms or hash signatures.

Content Analysis

Each Clavata.ai Policy can handle multiple content types,vand it's important to recognize that Rules may apply to different content based on slightly different criteria within the system:

Text content includes anything that involves written language, whether it’s a traditional text post or text overlaid on an image.
Image content refers to any visual files such as photos, drawings, or other non-text visuals.

While you may need to create specific Rules for targeting text and image content, the Policy analyzes all content types sent to it. For greater control, you might want to define separate Policies that evaluate different content types individually.

If you need more granular control, such as applying one Policy to each of your content types, Clavata.ai allows you to create separate Policies and use them to scan content accordingly.

Considerations

While Policies cover both text and image content, certain Conditions used to create Rules apply exclusively to text content:

The Exact Match Condition flags text that matches exactly with the words defined in the Policy. Exact Match is not case-sensitive and can handle repeated letters.
- For example, the Rule =("dog") will trigger an Exact Match if it appears as "Dog", "DOG", or even "ddoogggg" with repetitive characters.
The Fuzzy Match Condition allows for variations in the words to be flagged, such as common misspellings, leet speak, or other subversive techniques.
- For example, the Rule ~(“cat”) will trigger a Fuzzy Match if texts are detected, like “c4t

Exact Match Conditions do not apply to image content.

Examples

"drugs" [IN REFERENCE TO] "selling"

This Rule will flag:

text that falls in the context of selling drugs or drugs being sold.
Images where drugs are being sold

=(“dog”)

This Rule will flag:

Instances of the word “dog” in any case or character repetition. E.g., Dog, dOG, DOG, doggggg, ddddooogggg.
Not images of dogs of any kind, since this Rule is looking for an exact match and is relevant only to text.

Need more help? Contact our support team at support@clavata.ai