How to Report an AI Incident: A Step-by-Step Guide

Step 1: Capture the Evidence

The most important part of any incident report is verifiable evidence. Whenever you encounter an AI failure:

• **Screenshots**: Take clear screenshots showing both the prompt you entered and the complete response of the AI. Include model names or system details if visible.
• **Transcripts**: Copy the exact text of the conversation.
• **Files/Metadata**: Note down the date, time, and the specific application or API version you were using.

Step 2: Access the Report Form

Navigate to the [Report Incident](/submit) page. You can access this directly from the main navigation menu or by clicking the prominent "Report Incident" button.

Step 3: Describe the Incident

Fill out the report form with key details:

• **Title**: Write a brief, descriptive title summarizing what went wrong.
• **Description**: Explain what happened. What was the expected output, and what did the AI actually do? Mention any real-world impact or risk involved.
• **Category & Severity**: Classify the incident (e.g., Hallucination, Bias, Privacy) and select the appropriate severity level.

Step 4: Choose Your Privacy Settings

ALPAR AI respects your privacy. You can choose to publish your report:

• **Anonymously**: Your account details (if logged in) or identity will never be visible.
• **Publicly**: Your profile name will be associated with the report, building your reputation as an AI ethics advocate.

Before submitting, check the box to confirm you accept our terms and privacy policy. Our automated PII Guardian will automatically redact sensitive personal data like emails or phone numbers before storage.