Release Notes

If you encounter any issues or have questions about new features, please don't hesitate to contact our support team at [email protected]. We're here to assist you throughout the process.

November 2025

• Long Form Document Review: Introduced support for reviewing long-form documents with improved handling of multi-page content, better context retention, and optimized processing for large documents.

• Duplicate Feedback Prevention: Implemented automatic detection and prevention of duplicate task creation for identical feedback items, reducing noise and improving workflow efficiency.

• Expandable Table Rows: Added expand/collapse functionality for long question text in tables, improving readability and user experience when reviewing detailed policy questions.

• Shadow Model Testing: Released shadow mode capability for testing new AI models alongside production models without affecting user experience.

• Policy Engine Ordering: Improved policy execution ordering and management in the Policy Engine for more predictable and efficient policy application.

October 2025

• New Content Tab in Scanner: Introduced a dedicated "New Content" tab in the Content Scanner dashboard to easily view and filter newly discovered or changed content across all monitored sources.

• Task Manager Analytics: Added comprehensive analytics for task managers including regulatory actions tracking, project insights, and detailed metrics on review workflows.

• Regulatory Email Integration: Launched automated email monitoring for regulatory communications, enabling automatic ingestion and review of compliance-related emails.

• Assistant Use Cases: Expanded AI Assistant capabilities with new specialized use cases for regulatory research, policy analysis, and compliance workflows.

September 2025

• AI Assistant Tool: Introduced a new AI Assistant with comprehensive access to all Protege data, enabling intelligent queries and insights across your entire environment for enhanced workflow automation and data discovery.

• Shareable Document Review Links: Share document reviews with external users outside your organization using password-protected links. External user login data and access analytics are available upon request for compliance and audit tracking.

August 2025

• Real-time Asana Integration: Asana task changes are now pushed directly to Protege in real-time, eliminating sync delays and helping ensure your project management updates are immediately reflected across both platforms.

July 2025

• Teach your AI Approvals: All Teach your AI submissions now require approval in the Policy Engine before being used for model training to prevent misleading instructions from entering the finetuning process. You can edit or delete submissions during review, and optionally enable 3-day auto-approval for streamlined workflows.

• Notification Inbox: Access all your notifications in one centralized location with the new Notifications tab. View a complete history of Content Scanning policy alerts, Teach your AI updates, and reminders items.

• Daily Email Digest: Stay informed without constant interruptions. Enable daily business day email summaries to receive a consolidated overview of all notifications, policy feedback, and system updates delivered once per day during business hours.

June 2025

• Acknowledge Feedback: Users can now acknowledge feedback items to confirm they have been seen, without marking them as resolved—improving transparency and tracking of feedback that requires further review or action.

• Change Vision Models: Switching Vision Models is available upon request, message [email protected] or your support team to change it for any Policy step.

• Thumbs Up Feedback: Teach your AI that it's flagging the correct items.

• Display of Non-Visible Text: Webpages that are monitored by the Content Scanner now show a "Hidden Text" field. This field captures non-visible text — meaning text not shown to users directly (e.g., hidden via CSS, off-screen positioning, or within meta tags).

• Auto Risk Levels: Policy feedback can now be automatically classified as High, Medium, or Low based on LLM prompt-based definition and a configurable default level.

May 2025

• Default Brand Attributes: Use a Linked Database Sheet as a record for the default Attributes for all the Brands in your environment

• Content Scanner Dashboard Upgrade: The dashboard now includes richer metrics, including the ability to view and filter scan jobs by brand.

• Attachment Scanning Optimizations: Improved algorithms for calculating file similarity to reduce unnecessary repeat scans, helping lower client token usage.

• User Feedback Session Recordings: For error investigation and feature improvement, Protege may temporarily access video-based session click data leading up to a suboptimal or failed interaction. Recordings are automatically deleted within 7 days. To opt out, contact [email protected].

• Similar Usage: Users can now view all instances of similar phrasing across scanned pages and reviewed documents for any policy type (e.g., product comparisons, free stuff, legal qualifiers).—not just for claims. This helps teams identify recurring marketing language and help ensure all prior approved language is easily accessible.

• Real-time Sync Feedback for Tasks: When manually syncing tasks with Jira, Asana, or other integrations, you’ll now see real-time status updates directly in the dashboard.

April 2025

• Introduction of Llama 4 Maverick and Scout for analysis.

• Upgraded all large vision models to use Llama 4 for analysis.

• Mark as Reviewed in Content Scanner: Individual pages and full reports can now be marked as reviewed to track which pieces of content have been reviewed.

• Content Difference Analysis in Content Scanner: Website pages that have not changed since the last scan will now skip AI processing to save tokens. A link to the original scan of that page version will be shown.

• Performance increase: AI scans of documents have been optimized for load time.

• Change document attributes from within the Google Doc Extension.

March 2025

• Redesigned “Teach Your AI” Experience: Instruct your AI model to either skip or focus on specific types of feedback during document reviews.

• Nightly Agent Updates: Enterprise clients with customized ShieldLlama models now enjoy nightly model retraining for improved performance.

• New Analytics on the Agents Page: Monitor model accuracy and rollback models easily.

• Chrome Extension Helper Released: Speeds up Google Docs reviews with a dedicated helper extension.

• AI Explanations for Policies: Gain clear, detailed explanations for every policy decision.

• Resolved Feedback Section: A new section now shows AI feedback that has been resolved, ignored, or actioned.

February 2025

• Nightly Trainable ShieldLlama Base Model: Experience enhanced performance with daily retraining of your base model.

• Protege Sandbox Environment Launched: Enterprise clients can now test and experiment in a dedicated sandbox environment.

• Prohibited Content Policy Released: New guidelines have been introduced to enforce stricter content standards.

• General Policy Document Review: A streamlined review process for policy documents has been implemented.

January 2025

• Search across all Documents in your Protege environment

• Release of Archiving for long-term document retention with storage support for multiple clouds.

• Airtable Connectors for Attribute management.

• YouTube scanning for web monitor and social media influencer

December 2024

• Enhanced audio transcription: Upgrades to the audio transcript service for faster speed and improved accuracy.

• Improvements to Similar Claims accuracy: We upgraded the AI model used for generating Similar Claims. All existing claims will be migrated over without user action.

November 2024

• Support for Wrike Workspace and Asana Teams: Users can now add all projects from Wrike Workspace and Asana teams to Protege for streamlined project management.

• Skip Attachment Processing Mode: Introduced a new mode to bypass attachment processing for projects and workspaces, enhancing flexibility in data handling.

• Updated FinTech AI Policies: Enhanced FinTech AI policies to align with industry standards and address evolving compliance requirements.

• Document Page: Store files in the Document tab for AI testing, data labeling, and saving for long-term archiving.

• Custom Roles for Access Control: Added the ability to create custom roles for more granular access control permissions, empowering administrators with tailored user management.

• Typo Detection AI Policy: Released an AI-driven policy for typo detection in marketing brand and voice reviews to improve content accuracy and alignment with brand standards.

• Photoshop Asset Policy Review: Seamlessly review Photoshop assets for compliance with policies set in Protege AI, enabling efficient design and compliance workflows.

• False Positive Prevention - Negative Label Comparison: Improved handling of false positives by allowing better comparison of negative labels, reducing manual effort.

• False Positive Prevention - Cross-Medium Deduplication: Feedback across different mediums (e.g., audio and visual text) is now deduplicated, helping reduce the need to review similar text in multiple places.

• False Positive Prevention - Document-Level Exceptions: Set exceptions for specific documents, allowing the AI to ignore certain phrases or policies in context.

• False Positive Prevention - Brand-Level Exceptions: Apply exceptions across an entire brand for consistent handling of repetitive compliance issues.

• False Positive Prevention - Environment-Wide Exceptions: Extend exceptions to the entire Protege environment, enhancing flexibility in policy application.

• Subset Deduplication: When AI detects both a phrase and its containing sentence as independently triggering, Protege will now display the most appropriate feedback.

• Prompt Tuning (Assisted): Refine AI prompts with AI to improve accuracy and alignment with specific policy goals based on past labeled Exceptions, reducing false positives and negatives.

October 2024

• Non-Compliance Remediation Page: Added a new page to remediate non-compliance issues. Users can connect a project to Asana or Jira to track issues, assign tasks, and collaborate with cross-functional teams more efficiently.

• Query-Based Large Vision Models: Added query-based large vision models to identify specific compliance issues. Users can put a Prompt Identification step such as, “Identify any objects that look like a credit card, debit card, or prepaid card on this page.” to detect if there are payment cards on that page.

• Claims Approval Page: Added the ability to connect Asana or Jira task managers to the Claims Approval page for claim submission and approval processes, and streamlining the submission and storage within the Decision Store.

• Influencer Monitoring Tracking Page: Introduced a new page to monitor influencer content across social media platforms for compliance with legal and brand guidelines. Automatically track posts, detect non-compliant content, and link to Asana or Jira for seamless issue tracking and resolution. Enhance campaign oversight with real-time alerts and detailed reporting on influencer activities. With this page, you can help ensure your content complies with advertising guidelines, such as disclosure requirements, FTC regulations, and platform-specific policies.

September 2024

• Facebook Page Scanner: Introduced a new scanner to automatically monitor and analyze content on Facebook pages for policy violations and compliance monitoring.

• Apple App Store Listing Page Scanner: Added a new scanner to automatically review app descriptions, images, and metadata for policy compliance on Apple App Store listings, providing detailed reports on potential issues.

• Android Play Store Listing Page Scanner: Launched a new scanner to automatically review app descriptions, images, and metadata for policy compliance on Android Play Store listings, providing detailed reports on potential issues.

• Scanner Reports and Jobs Enhancements: Added new sorting and filtering options on the Scanner Reports and Scanner Jobs pages, allowing for more efficient navigation and management of scan data.

• Shareable Filtered Links: Introduced the ability to share applied filters on the content scanner with teammates via shareable links, making collaboration and data review more streamlined.

• Scan Report Naming: Introduced the ability to organize scan results more efficiently by assigning custom names to one-time and recurring jobs, streamlining result management and retrieval.

• Multi-Variable Attributes: Added support for assigning multiple attribute values to each piece of reviewed marketing, allowing for more detailed and flexible classification and analysis.

August 2024

• Microsoft Word 365 Add-In: Enhanced speed, with noticeable reductions in the time required to display the first result and show all results.

• Google Extensions: Enhanced speed, with noticeable reductions in the time required to display the first result and show all results.

• Open-Source Large Language Models: Access new open-source large language models to execute your Steps faster.

• Attributes Feature Support: Now supports the creation of new attributes for Tasks. You can set Attributes in Settings and designate text languages (e.g., disclosures) from a Google Sheet. Protege AI will automatically adjust the execution of Policies and Steps for each Task based on the assigned Attributes.

• Multi-Language Document Translation and Review: We now support AI translation of documents in languages other than English, allowing for quick human and AI Policy review.

• Copy Policy: Duplicate a Policy and add it to any Policy Group in your environment.

July 2024

• On the Task Page, a “Change Document Evaluation Type” selector has been added. Changing this option will automatically rescan all attachments for the Task.

• The Task page now shows Activity and Comments from linked Jira or Asana tasks.

• We’ve improved zoom in/out features and added a 100% size button for viewing Image and PDF files.

• You can now filter Decision Store records by Brand.

• You can export Trademark and Claim records from a Decision Store to a CSV file.

• Brand Selection is now available during Document Reviews. You can choose a brand when reviewing documents from Google Workspace, Microsoft Office, and the Figma extension.

• The Figma extension now supports scanning a single selected Figma frame instead of the entire page.

• Wayback Mode lets you see all Scans over time for a specific web URL being monitored in Content Scanner. You can quickly navigate to other scanned pages on that domain.

• Protege AI now automatically adds comments with QA review links in new Jira and Asana Tasks.

• Policy Feedback titles can now be customized along with new highlighting colors, and changes can be made from the Label field in the Policy Engine.

June 2024

• Content Scanner now supports LinkedIn and X.

• Risk Scores (High, Medium, and Low) are available for Trademarks, Claims, and Policy Feedback Cards.

• Actions - a series of automated steps that can be triggered after a document is analyzed by Protege AI. The initial launch includes:

  • Sending all AI Feedbacks above a customer-assignable risk threshold to Asana and Jira.

  • Approving a marketing review request if the documents are below a certain risk threshold.

  • Creating a new Asana or Jira ticket with a list of issues to fix if Content Scanner results are above a certain risk threshold.

May 2024

• Automated document review integration with Microsoft Office 365 SharePoint when files are linked in Asana and Jira tickets.

• Automated document review integration with Frame.io when review links are attached in Asana and Jira tickets.

• Enhanced sorting and filtering options on the Task list page.

• Zooming and panning capabilities for PDF and image attachment previews.

• Ability to comment from Protege AI to Asana, Jira, and Frame.io.

• Detailed explanations for Policy Engine actions.

• New Protege AI plug-in for Figma.

• Self-service creation for one-time and recurring Web Scan jobs.

April 2024

• Secure OAuth login for Jira Task Manager Review.

• New integrations with Asana, Wrike, and Smartsheets for automated ticket review.

• Automated document review for linked Google Docs and Slides in Jira and Asana tickets.

• Microsoft Word 365 Add-In for Protege AI.

• UX enhancements to the Protege AI Dashboard.

• Introduction of approval audit reports.

March 2024

• Support for up to 100-page Google Slides and very long Google Docs.

• New UX for AI feedback in the Google Workspace sidebar.

• Dashboard bug fixes.

• Improved admin account permissions, user roles, and user deletion options