About AI Check
Overview
The AI processing of this feature uses Microsoft's "Azure OpenAI Service" (hereinafter "OpenAI").
To use this feature, a separate contract for that service is required on the customer's side.
Also, to use the AI Check feature, please contact the support desk in advance for the OpenAI "API key".
AI Check is a feature used by the AI Check action of a policy.
When sending an email, it checks the email content with AI (Azure OpenAI API), and when the AI judgment level is within the range of the warning detection level, it prompts the user to recheck the content on the Send Alert screen of the User Outlook Add-in.
The administrator can enable AI Check and configure the basic settings and prompt configurations for each settings information (tenant / group / user) and apply them to users.
AI Check Result Items
The following items are displayed as the AI Check result. Each item corresponds to the evaluation item keyword set in the prompt.
| Item | Description | Prompt evaluation item keyword |
|---|---|---|
| AI judgment level (10 levels) | Displays the AI judgment level set in the prompt on a 10-level scale. (1 = no problem, 10 = highly inappropriate, etc.) | level |
| AI judgment type | Displays the type of AI judgment set in the prompt. (For example, in the case of "harassment", "sexual harassment", "power harassment", etc.) | kind |
| Cause | Displays the expression or context in the email body that the AI used as the basis for the above judgment. | reason |
| Reason | Displays the reason (background or intent) why the AI made the above judgment. | because |
| Advice from AI | Displays improvement suggestions and advice for making the text better. | advice |
The AI Check result is generated by AI and is not necessarily accurate information.
Settings Required to Use AI Check
| Setting | Description |
|---|---|
| Basic Settings | Configures the connection information and parameters for Azure OpenAI. For details, refer to Basic Settings. |
| Prompt Configuration | Registers the prompts that serve as the AI Check perspectives. For details, refer to Adding a Prompt Configuration. |
| Policy | Set AI Check as the policy action, and specify the prompt to run and the warning detection level. |
Enabling the "administrator manages" setting for each configuration prohibits changes in the User Outlook Add-in and allows distributing the content configured by the administrator.
If you set up access restrictions on the OpenAI side, you must allow the IP addresses of 365 Alert OAI.
For the IP addresses required for the allow settings, contact the support desk.