Skip to main content

​
20260304

​
1. Entity Library Enrichment

Users can now enrich entities directly from the Entity Libraries during privilege review.
  • Select entities detected in documents and enrich them with information from existing libraries.
  • Save newly identified entities directly to the library for future reuse.
  • This improves consistency across matters and strengthens privilege detection by leveraging curated entity knowledge.

​
2. Redaction Prediction and Privilege Waiver Detection

Identifying potential redactions and privilege waiver risks within email threads.
  • Flags possible privilege waiver scenarios when privileged communications are shared with third parties within an email chain.
  • Helps reviewers identify risk areas earlier in the review process.

​
3. Privilege Inference Logic

Privilege inference logic has been refined to align with model outputs and legal review expectations.
  • Attorney and third-party involvement are now treated as strict inference rules.
  • Improves consistency between automated inference and LLM classification results.

​
4. Family-Level Privilege Inference

Parent documents marked as privileged now properly propagate privilege status to related family documents where applicable.

​
5. Name Normalization Formats

Move name format operations to be within Name Normalization and add more name format templates.

​
20260115

​
1. Library Sync & Entity Labeling

We have optimized the workflow for identified entities (People, Organizations, etc.). When an entity exists in the verified libraries, the system now automatically syncs and copies tags without additional changes.
  • What’s New: Eliminates additional LLM inference for known entities.
  • Key Benefit: Ensure tag consistency when entities are already “known” to the system.

​
2. Enhanced Name Reference Detection

The system can identify individuals based on common naming conventions.
  • What’s New: Improve detection for nicknames and name variances.
  • Example: If the system identifies “Nicholas” as an attorney, it will now automatically detect and flag “Nick” as the same entity.
  • Benefit: Reduces manual alias mapping and ensures more comprehensive privilege flagging.

​
3. Improved Recall Categories (Borderline Detection)

To ensure we aren’t missing “gray area” documents, we’ve introduced two new high-recall categories. These function similarly to the logic used in aiR.
  • New Categories:
    1. Legalese (No Priv): Captures documents with heavy legal jargon that may not be privileged but are relevant for review.
    2. Attorney Involvement (No Priv): Flags documents where an attorney is present but the content doesn’t immediately trigger a “Privileged” status.
  • Benefit: Acts as a safety net to catch borderline documents that previously might have been missed, ensuring a more defensible review process.

​
4. Automated Family Log Generation

Providing context for attachments is now automated, making privilege logs much more descriptive and easier to navigate.
  • What’s New: Automatic generation of relationship descriptions.
  • Example: For an Excel sheet attached to a flagged email, the system will generate a description like: “Attachment to a privileged email.”
  • Benefit: Streamlines the creation of privilege logs and provides immediate context for why a non-text document was withheld or flagged.

​
5. Protocol-to-Workflow Integration

We have introduced a direct pipeline to import instructions from a Privilege Review Protocol directly into a Workflow.
  • What’s New: Users can now import specific privilege instructions and tags from a Protocol document. The system then automatically parses these instructions to sync existing tags or create new ones.
  • Key Benefit: Eliminates manual data entry and “human error” when setting up tag structures.

​
20251223

We’ve updated the labs instance with several new features and improvements to support privilege workflows, review/QC, and usability. Highlights below:

​
1. Privilege Classification Output Enhancements

  • Added “Privilege Prediction” and “Privilege Ground Truth” columns to the privilege classification output.
  • These store binary decisions (Privileged / Not Privileged) to make filtering, sorting, and downstream workflows (review, QC, batching) easier.

​
2. Thread Output Logic Improvements (Privilege Inference)

  • Updated the thread light-switch logic to better reflect real-world privilege rules:
    • When a third party is introduced, all prior messages in the thread are marked Not Privileged.
    • When an attorney is introduced, all subsequent messages in the thread are marked Potentially Privileged, unless there is a clear privilege break.
  • This improves consistency and interpretability of thread-level privilege decisions.

​
3. Save Workflow as Template

  • Added a “Save as Template” button to workflows.
  • Users can now save a customized workflow and reuse it as a template for future runs.
  • This makes it easier to standardize workflows and repeat common setups.

​
4. View & Download Historical Runs

  • Added a “View Runs” button on the flow results page.
  • Users can:
    • View prior runs of the same workflow
    • Download results from historical runs
  • This improves auditability and review flexibility without re-running workflows.

​
5. Create Review Batches from Privilege Results

  • Added a “Create Review Batch” button on:
    • Privilege Classification
    • Privilege Log Generation pages
  • Users can now create review/QC batches directly from AI predictions.
  • This bridges AI outputs directly into human review workflows.