Startup Securiti Inc. and Databricks Inc. today announced a partnership focused on helping joint customers develop artificial intelligence applications.

The collaboration will see Securiti integrate its Gencore AI tool with two components of Databricks’ popular cloud data platform. The first is the Delta table data format, while the other is the Mosaic AI toolkit for building AI software.

San Jose, California-based Securiti is backed by more than $150 million in funding. Its flagship product is a data protection platform called the Data +AI Command Center. The software promises to both protect business records from hackers and ensure that they’re stored in accordance with regulations, two tasks which historically had to be performed using separate tools.

Last year, Security expanded its product portfolio with a tool called Gencore AI. The software helps AI applications access the data that they require to answer user prompts. Additionally, it ensures the information retrieval is done in a secure manner that complies with privacy laws.

Alongside Gencore AI, Securiti’s new partnership with Databricks focuses on the latter company’s Delta table and Mosaic AI features.

Databricks provides a data lakehouse platform that enterprises use to store and process business information. It can hold structured, semistructured and unstructured records. There’s built-in support for ACID, a reliability standard that ensures data isn’t lost in the event of technical issues. 

By default, Databricks stores structured information in data structures called Delta tables. Customers can make those Delta tables accessible to AI models with Mosaic AI, a suite of machine learning features built into the platform. The latter offering is based on assets that Databricks obtained through a $1.3 billion startup acquisition. 

Thanks to the partnership announced today, companies can now use Gencore AI to stream records from internal systems into Databricks. A new integration between the products makes it possible to turn the imported information into Delta tables and then make it accessible to AI models using Mosaic AI.

According to Securiti, Gencore AI can process data from hundreds of sources including cloud applications and on-premises infrastructure. The platform automatically removes duplicate items from the information being moved. Additionally, it filters sensitive data such as credit card numbers that can’t be used for AI training.

Once a dataset is in Databricks’ platform, companies can use it to train AI agents. Those are AI applications that perform business tasks with little to no manual input. The built-in Mosaic AI toolkit also eases related tasks such as equipping neural networks with RAG features. 

After a new AI application exits the development phase, customers can use Gencore AI to secure it.

The systems from which an AI application retrieves data often have access controls in place. For example, an accounting tool may only allow members of the accounting team to edit financial records. Gencore AI carries over those data access restrictions to the AI applications in which the records are used.

Gencore AI can also block AI-specific cyberattacks. That includes prompt injection attacks, which use malicious instructions to trick an AI model into performing tasks it was not designed to perform. Gencore AI can similarly block harmful AI output and irrelevant prompt responses.

 Image: Databricks