A visual content map and the Train-Test-Deploy-Analyze loop

June 1, 2026

New Features

• Quality control reports can now be exported as a CSV file directly from the details page.
• You can now create a knowledge task directly from the quality control report when the AI is unable to answer a question, addressing the gap on the spot.
• The content library now includes a visual, node-based content map for easier navigation.
• User details now include a direct link to the data erasure request form for privacy compliance.
• When the AI answers a question, a “Read more” button now appears (if your workspace has the content library enabled) that opens a full article view on the subject.

Improvements

• The navigation structure in Unless has been updated to provide a more straightforward workflow, organized around the four phases of the loop: Train → Test → Deploy → Analyze.
• The browser extension now displays a helpful error message instead of going blank when something unexpected happens.
• The AI’s sensitivity to potential misuse attempts has been tuned to reduce unnecessary blocks on legitimate questions.

Bug Fixes

• Reporting a flagged response works as expected again.
• The browser extension’s authentication flow is working correctly again, and website access checks have been re-enabled.