New self-serve capability lets users train custom entity detection models on their own data—no data science expertise required.
San Francisco, CA — November 17, 2025 — Tonic.ai, the leader in privacy-preserving data generation and transformation for AI development, today announced the launch of Custom Entity Types, a new feature within Tonic Textual that empowers users to build custom entity detection models on their own data, within their own infrastructure or on Tonic’s secure cloud.
The feature introduces a fully self-serve workflow for defining and training custom entities—enabling customers to improve or extend Textual’s detection capabilities to serve industry and organization-specific text through an approachable UI. Using large language models (LLMs) for assisted annotation and model distillation, Textual makes it easy for organizations to create high-accuracy, domain-specific models that adapt to their unique data.
A self-serve breakthrough for sensitive text data
“Custom Entity Types puts the power of model customization directly in the hands of our users,” said Adam Kamor, Co-founder and Head of Engineering at Tonic.ai. “Our customers can now train models and define unique entities themselves—achieving the necessary level of detection and confidence within their workflows—even with highly nuanced text, and without bringing in data science resources.”
Organizations in highly regulated industries such as healthcare, financial services, and legal technology face growing pressure to de-identify sensitive text data while preserving accuracy and context. Custom Entity Types addresses that need by combining a self-serve interface for annotation with private model training, allowing for bespoke detection that’s unique to their specific use case.
How it works
With Model-Based Custom Entities, users can:
- Upload documents containing examples of the desired entity type.
- Leverage LLM-assisted annotation to automatically identify potential entity spans.
- Review and refine annotations through an intuitive interface.
- Train a custom entity detection model on their labeled data.
- Deploy the model securely within their environment for real-time use.
Because the model is trained on the customer’s own data, it achieves exceptional precision and recall for domain-specific entities—whether that’s prescription names and biometric data for healthcare organizations—or unique account information for financial services organizations.
Accelerating onboarding and adoption
The new capability also accelerates evaluation and onboarding for new customers. Instead of waiting for custom models to be developed by Tonic’s internal team, users can now generate their own entity models during product evaluation—reducing time-to-value and improving adoption rates across enterprise deployments.
“Custom Entity Types not only improves model accuracy—it makes AI data privacy more accessible,” said Whit Moses, Senior Product Marketing Manager at Tonic.ai. “By putting model training in the hands of the user, we’re eliminating a key bottleneck to responsible AI innovation.
Visit the blog for a deeper dive into this new capability. Teams working with unstructured data can try Textual for free or book a demo with an expert at Tonic for a deep dive into the product as it pertains to their specific use case.
.avif)
