Today marks an exciting milestone for us here at Tonic.ai as we unveil our latest innovation to help enable data privacy for unstructured text data at scale: the Tonic Textual Snowflake Native App, designed specifically for organizations developing machine learning training workloads in Snowflake, who need advanced, seamless, and secure data redaction and synthesis. We are excited to partner with Snowflake to bring the powerful data protection capabilities of Tonic Textual directly into the hands of the data engineers, data scientists, and decision-makers in Snowflake’s extensive user base.

Let’s dive into the benefits this offers to anyone building analytics, testing, and AI workloads in Snowflake today.

What is Tonic Textual?

Tonic Textual is our enterprise platform for AI development, designed to eliminate integration and privacy challenges ahead of RAG ingestion or LLM training—two of the biggest headaches for development teams trying to move quickly on AI initiatives. After pointing Textual at any common cloud object storage, it will automatically parse most imaginable files types, extract clean text, transform it into a standardized format, tag the data with entity metadata tags, and provide optimal chunking if desired. Textual utilizes proprietary NER models trained across a variety of formats, contexts, and domains to meticulously identify entities. Once detected, our application can also redact and tokenize any sensitive entities found, optionally replacing it with high-quality synthetic data that retains the original data’s semantic realism and functional utility. This means that your data remains usable for analytics, data pipelines, model training, and LLM app development—all while ensuring that your sensitive information remains just that: sensitive and private.

Imagine being able to seamlessly tame the chaos of unstructured data into a form ready for analysis, ML model training, and LLM app development without the headache of building data pipelines and risk of exposing sensitive data—Tonic Textual makes this a reality!

Introducing the Tonic Textual Snowflake Native App

Tonic has been working behind the scenes with Snowflake for months as a strategic partner to leverage the Snowflake Native App Framework’s private preview integration with Snowpark Container Services, which enables us to deploy Tonic Textual as a containerized application directly within your Snowflake environment. The Tonic Textual Native App features two main components: a detection service and a redaction service. When you run the TEXTUAL_REDACT function, text data in your Snowflake table is processed and a new privacy-preserving result-set is returned with redacted or synthesized text. Under the hood, our proprietary NER models are springing into action, identifying sensitive information which is then replaced with redacted or synthesized versions. It also utilizes Snowflake’s new GPU compute pools for exceptional performance at Snowflake scale. What makes this especially cool? Your data never leaves Snowflake, leveraging Snowflake’s robust security posture to maintain your data’s confidentiality and integrity.

Setting up Tonic Textual is a breeze. A few commands to set up a compute pool and permissions, and that’s it—you’re all set to start redacting text right within your Snowflake workflows. It’s as simple as clicking “Install” from the Snowflake UI, and the app is ready to go. You control who accesses the app and how it’s used, all with minimal setup fuss.

The integration of Tonic Textual as a Snowflake Native App is a game-changer. Here’s why:

1. Seamless integration: Say goodbye to complicated setups and configurations. With Tonic Textual, you can get up and running with just a few clicks directly from the Snowflake Marketplace. This ease of installation means you save time and reduce the risk of errors, letting you focus on what really matters—your data.
2. Stay within the ecosystem: All processing happens within Snowflake’s secure environment. This containment ensures that your data does not traverse multiple platforms, thereby reducing the risk of breaches and leaks.
3. Unified billing: Using Tonic Textual native to Snowflake allows for unified billing. This integration simplifies your subscriptions and financial processes, making life easier for your budgeting and finance teams.

Use cases of Tonic Textual

Data privacy and compliance: Safeguarding sensitive information is paramount. Tonic Textual helps organizations meet stringent compliance requirements for data privacy laws such as GDPR, HIPAA, and CCPA by effectively de-identifying sensitive data.

Preventing data leakage in AI: When training AI and machine learning models, preventing the model from memorizing sensitive data is essential to avoid unintended data leakage. Tonic Textual addresses this by replacing real data with synthetic alternatives that maintain semantic realism without compromising privacy.

Secure lower environment testing: Before deploying data into development, testing, or staging environments, it is crucial to ensure that sensitive data is not exposed. Tonic Textual facilitates this by de-identifying text data before it is sent to lower environments. This not only protects your data but also enables secure analytics, data pipeline development, machine learning model training, and LLM development.

Our customers are already finding innovative ways to use Tonic Textual across various sectors:

• Healthcare: Ensuring PHI is protected while analyzing patient data to improve treatment outcomes and building LLM chatbots to streamline healthcare worker knowledge access and productivity.
• Financial Services: Safeguarding PII while maintaining the richness of datasets for fraud detection.
• Retail: Enhancing customer service through Tonic Textual’s ability to synthesize and redact sensitive data, enabling the development of superior customer service chatbots and comprehensive analytics on customer service interactions. This ensures that each interaction is personalized and privacy-compliant, contributing significantly to customer success.

The importance of data privacy and stewardship

In today’s digital age, data privacy is not just a necessity but a mandate. Regulatory frameworks like GDPR and HIPAA have set the bar high for how data should be handled, emphasizing the importance of protecting or minimizing personal and sensitive information in your data stores. Tonic Textual empowers organizations to uphold these standards, ensuring that as your data is utilized for growth and innovation, its integrity and confidentiality are never compromised.

Good data stewardship is about more than compliance; it’s about ethical responsibility. By de-identifying data before it is used for real-world production use cases—like training ML models—you protect not only your customers’ privacy but also the reputation and reliability of your business.

A corollary for AI

The rise of AI and machine learning models demands more data than ever. However, the use of sensitive data in training these models can lead to unintended consequences like model memorization, where the model inadvertently learns to replicate sensitive information, posing significant privacy risks.

Tonic Textual ensures that your training datasets are cleansed of sensitive data, enabling the responsible development of AI technologies. With our tool, models learn from data that maintains real-world complexity without compromising privacy.

Join Tonic.ai and Snowflake in this new wave of AI 

Tonic.ai is on a mission to transform how businesses handle and leverage sensitive data while still enabling developers to do their best work. We invite you to join us on this path, ensuring that our data drives us forward, while remaining secure and private every step of the way. Together, we’ll set new standards for data privacy and security. Stay tuned, stay secure, and let’s keep pushing the boundaries of what our data can do for us—safely and responsibly!

You can learn more about Tonic Textual on our website and see how easy it is to keep your data secure and useful. Tonic Textual will be available on the Snowflake Marketplace in early June, 2024—if you’re interested in being an early beta user, let’s chat.