Oracle Corp. today announced the general availability of Oracle AI Database 26ai and Oracle Autonomous AI Lakehouse, both aimed at supporting artificial intelligence training and inference across cloud and on-premises environments.

Oracle said the releases continue its strategy of integrating artificial intelligence capabilities directly into its core data platforms. Oracle AI Database 26ai is a long-term support release that replaces the current Oracle Database 23ai. The update embeds AI functions natively in the database engine to support applications without requiring separate data movement or third-party tools.

Among the capabilities in the new release are AI vector search, Model Context Protocol server support and a framework for creating and deploying AI agents inside the database. Oracle said the features allow enterprises to combine private data with public information and enable multi-step AI reasoning within the database.

The database supports integration with a variety of popular tools and models, including Open Neural Network Exchange for representing machine learning embedding models, open agent frameworks and widely used large language models. It’s also compatible with the Apache Iceberg table format for data lakes.

Security enhancements include quantum-safe encryption for both data-at-rest and data-in-flight using government-approved algorithms. Oracle said this level of encryption is needed to safeguard data against future threats from quantum computers, which can unscramble existing cryptographic algorithms quickly.

Hardware acceleration is available via Oracle Exadata for AI, a high-performance database platform for AI workloads. Other performance-enhancing features include remote direct memory access and tiered storage. Oracle’s Private AI Services Container enables AI model instances to run on a customer’s own infrastructure without exposing data to external providers. Support for future Nvidia Corp.-based acceleration is also planned.

For developers, Oracle is introducing a no-code AI agent builder called AI Private Agent Factory and a natural language-driven application development environment within the company’s Application Express low-code development platform. The tools are intended to support faster development of applications and workflows using in-database agents.

Other features planned for Oracle AI Database 26ai include a unified data model for relational, JavaScript Object Notation and graph data, data annotations for improving AI model output, a globally distributed database model, True Cache for automatic transactional data caching, and a SQL firewall to block unauthorized activity.

Existing users of Oracle Database 23ai can apply the update at no charge. Oracle said no application re-certification or database upgrade is required.  and that AI Vector Search is included at no additional charge.

AI lakehouse

Oracle also introduced Autonomous AI Lakehouse, a new platform that combines Oracle’s Autonomous AI Database with the open Apache Iceberg table format. The platform is available on Oracle Cloud Infrastructure, Amazon Web Services Inc., Microsoft Corp.’s Azure, Google LLC’s Cloud and Oracle’s Exadata Cloud@Customer.

The Autonomous AI Lakehouse is designed to support AI and analytics workloads across cloud and hybrid environments while maintaining compatibility with tools such as Databricks Inc., Snowflake Inc. and AWS Glue. Oracle said the platform removes the operational burden of moving data.

Key capabilities include support for AI vector search on Iceberg tables, JSON-relational duality, property graph analytics and Select AI, which converts natural language into SQL queries. The features are intended to let users perform complex searches and analysis across structured and unstructured data.

A Data Lake Accelerator, currently in limited availability, dynamically scales network and computing capacity based on query demand while billing only for the resources used.

A new Autonomous AI Database Catalog aggregates metadata and data sources across various clouds and catalogs. It’s intended to simplify data discovery and governance across heterogeneous platforms.

Other features include Exadata Table Cache for caching Iceberg data in flash storage, GoldenGate for Iceberg for real-time data streaming into Iceberg targets and a table hyperlink mechanism to enable temporary access to shared data without managing complex permissions.

Both Oracle AI Database 26ai and Autonomous AI Lakehouse are now available.

Photo: Flickr CC