Compliance Agent [Beta]

This release introduces Compliance Agent [Beta], a new feature that generates responses to compliance questions using your organization's documentation. Compliance Agent [Beta] will be made available to all customers and vendors as part of this release, providing an early opportunity to explore how AI can assist with compliance response workflows. As a beta feature, it is being released in an initial form and will be refined over time.

Description: Compliance Agent [Beta] generates responses to compliance questions based on your organization's documentation.
Availability: This feature will be available to all customers and vendors as part of this release.
Notes: Compliance Agent [Beta] is being introduced in an initial form and will be refined over time.
Access: Compliance Agent [Beta] is available within the Compliance Framework workflow when answering questions.

Limitations in Beta

Compliance Agent [Beta]

Only available for the NIST AI RMF Starter Pack v2 Compliance Framework.
No available for existing NIST AI RMF Starter Pack v2 assessments (in-progress or completed).
To use: add a new NIST AI RMF Starter Pack v2 to an AI System and open the framework.
Uses only embedded PDFs.
Avoid duplicate content (same document, different filename) - it can limit results.
Responses take ~2 minutes per question; running across an entire framework will take longer.
Accuracy and quality of responses depends on the quality of the documentation available in your environment.

Embedding

Only newly uploaded documents are embedded (historical documents will not be processed in this first release).
PDF-only; text-only (images/diagrams inside PDFs aren't used).
Password-protected PDFs aren't embedded, even if upload succeeds.
Existing document library file size limits (20MB) still apply.

Bug Fixes & Other Improvements

In addition to this new capability, the release includes a series of quality and reliability improvements. These updates address issues related to Compliance Scoring, status filtering in the Bill of Materials list, model details in Arena, and exporting Detect AI results. Together, these changes are intended to improve platform consistency and overall user experience.

Compliance Scoring: Fixed an issue where final results did not display after all questions were scored.
Bill of Materials List: Corrected status filtering to return complete and accurate results.
Arena (Model Details): Removed references to deleted AI Systems; deleted systems no longer appear.
Detect AI: Resolved CSV export failures for large result sets. Corrected duplicate messages preventing the completion of larger scans.

Did this answer your question?