Tonic Textual on Microsoft Fabric: Now in private preview

Enterprises want to use real-world, unstructured data to inform AI initiatives, RAG systems, and model training; but privacy concerns often block progress – especially in the face of compliance requirements and the risk of data leakage. With Tonic Textual now available in private preview on Microsoft Fabric, teams can transform sensitive text and documents into AI-ready, compliant datasets—directly inside Fabric.

Microsoft Fabric is an end-to-end analytics and data platform designed to unify how enterprises ingest, manage, and operationalize data for downstream use cases. For AI teams, Fabric provides a central place to prepare data, connect to compute environments, and integrate with large-scale modeling pipelines. By consolidating storage, governance, and analytics tools into a single environment, Fabric helps teams reduce friction and accelerate AI adoption while maintaining enterprise-grade security and scalability.

Tonic Textual makes sensitive text safe for AI development 

Tonic Textual de-identifies unstructured data — from raw text and chat logs to PDFs, Word docs, text within images, and audio so teams can safely use it for AI/ML without exposing personal information (PII and PHI) contained within.

• Detect, redact, synthesize, or tokenize: Out-of-the-box detection with custom detectors for unique and organization-specific entities.
  
• Maintain relationships across datasets: Consistent mapping of entities to the same synthetic value across files to preserve relationships and analytics integrity.
  
• Leverage important files: Works across the most important file types including PDFs, DOCX, images containing text, audio.
  

Create model-safe data: Replace sensitive content with tokens or realistic synthetic values to avoid leaking PII into model weights.

“You can’t risk putting customers’ PII/PHI into the weights of the model you’re training… so you run your data through Tonic Textual to generate synthetic unstructured data for training.”

Adam Kamor Co-founder and Head of Engineering

Why Tonic Textual + Microsoft Fabric (Better Together)

Placing Tonic Textual upstream in Fabric pipelines converts sensitive inputs into de-identified, realistic outputs that are safe to use across Fabric’s AI services.

“Data is ‘radioactive’ until you deal with the sensitive information inside it… having Tonic at the beginning of your pipelines makes a ton of sense.”

Roy Hasson Sr. Director of Product

What you get:

• Frictionless ingestion → safe outputs: Use Fabric’s 200+ connectors and Lakehouse to land data; point Tonic Textual at source folders and an output Lakehouse for sanitized copies.
  
• End-to-end privacy by default: Remove/replace PII before it flows into notebooks, Spark, warehouses, AI Foundry, or RAG systems.
  
• High-fidelity synthetic data: Keep structure, semantics, and cross-document consistency so downstream analytics and training remain meaningful.
  
• Lower governance burden: Reduce reliance on complex, late-stage permissioning; handle risk before proliferation.
  

Now in Private Preview

Early adoption comes with the chance to help shape the roadmap of the integration. Early participants will work directly with Tonic’s engineering team, providing feedback and influencing how the product evolves to meet enterprise-scale needs.

Teams can get started by requesting access here. Once submitted, a member of the Tonic.ai team will schedule time to learn about your use cases and confirm eligibility before provisioning access. Seats in the preview are limited, so early enrollment is encouraged for organizations that want to both gain a head start and influence the future direction of this capability. Teams best suited to participate in this opportunity meet the following criteria: 

• Currently storing data in Microsoft OneLake 
• Working with unstructured text (or audio) that contains PII or other sensitive information 
• Seeking to leverage this data for AI initiatives and model training 

The private preview is open now. If you’re working in Microsoft Fabric and want to confidently bring sensitive text into your AI workflows, this is your opportunity to get hands-on with the integration before general availability.