Managing Keyword Policies

Keywords, also known as user-defined secrets, let you monitor custom words or phrases that the Honest AI Shield should watch for and intercept during your AI conversations.

There are two types of keyword policies:

1. Sensitive Keywords (for your messages)

These are words the shield will look for in messages you send. If found, the shield will alert you and offer to redact them.

Example use case: Your passport number is "PB1234567" and you want to make sure you never accidentally mention it to an AI.

How to set up:

1. Go to the Policies tab
2. Expand "Manage Keywords Policies"
3. Click "Sensitive Keywords" (labelled "in prompts")
4. Type your keywords in the text box, one per line
5. Click Save

2. Censored Keywords (in AI responses)

These are words the extension will look for in AI responses. If found, the response will be blocked.

Example use case: You are doing research and want to block the AI from showing competitor brand names.

How to set up:

1. Go to the Policies tab
2. Expand "Manage Keywords Policies"
3. Click "Censored Keywords" (labelled "in responses")
4. Type your keywords in the text box, one per line
5. Click Save

warning

Censored keywords in AI responses only take effect when the AI Response Moderation toggle is turned on in the Shields tab. Without it, only your outgoing messages are scanned for sensitive keywords.