Oracle unveils AI Database 26ai with advanced security

Oracle has announced the release of Oracle AI Database 26ai, a major update that introduces artificial intelligence integration across its flagship database platform.

The updated offering incorporates AI into the database's architecture, targeting both operational and analytical workloads across multicloud and on-premises environments. The company stated this update is intended to advance Oracle's approach of bringing AI capabilities directly to where enterprise data resides.

With this release, Oracle is focusing on enabling customers to utilise AI for a wide span of data types and development processes. The platform supports capabilities such as AI Vector Search, AI-enhanced database management, application development, and analytics. These integrated functions are designed to facilitate the running of advanced agentic workflows, which allow organisations to combine proprietary data with external public information for complex queries and actions.

Oracle AI Database 26ai is a long-term support release, replacing Oracle Database 23ai. Users can transition by applying the current release update, without the need for complete database upgrades or application re-certification. 

A key feature of Oracle's strategy is an open approach that incorporates industry standards and compatibility with various AI systems. Oracle AI Database now includes built-in support for the Apache Iceberg open table format, the Model Context Protocol (MCP), major large language models (LLMs), agentic AI frameworks, and the Open Neural Network Exchange (ONNX) embedding models. The database can be deployed across multiple environments, such as Oracle Cloud, leading hyperscale clouds, private cloud, and on-premises infrastructure.

On the security front, Oracle AI Database 26ai implements quantum-resistant encryption standards. It uses NIST-approved ML-KEM algorithms to secure data during transit and maintains its support for quantum-resistant encryption at rest. This approach seeks to safeguard organisational data against future risks associated with quantum computing-based attacks.

One of the headline additions with this release is the Oracle Autonomous AI Lakehouse, which supports the Apache Iceberg open table format for enterprise-wide analytics. The Lakehouse is available across Oracle Cloud Infrastructure, Amazon Web Services, Microsoft Azure, and Google Cloud, and is interoperable with Databricks and Snowflake in those environments. It delivers performance via Exadata and offers serverless scalability on a pay-per-use basis, allowing organisations to leverage AI for data stored in external data lakes without moving it.

"By architecting AI and data together, Oracle AI Database makes 'AI for Data' simple to learn and simple to use," said Juan Loaiza, Executive Vice President, Oracle Database Technologies at Oracle. "We enable our customers to easily deliver trusted AI insights, innovations, and productivity for all their data, everywhere, including both operational systems and analytic data lakes."

The new AI Database also introduces founding technologies such as unified hybrid vector search, which integrates vector search with relational, text, JSON, knowledge graph, and spatial searches. This allows for complex retrieval tasks across multiple data forms, enabling AI-driven insights to be generated from documents, images, video, audio, and structured datasets.

The MCP Server Support feature grants AI agents powered by LLMs access to databases for iterative, multi-step reasoning tasks. Built-in data privacy controls offer protection at the row, column, and cell levels, with dynamic data masking and granular access controls, aiming to enable AI use while safeguarding sensitive information.

Further, Oracle Exadata for AI delivers AI acceleration at scale by leveraging hardware and software co-design, offering vector query acceleration, elastic scaling, and high throughput for demanding workloads. Security is enhanced by the Oracle Private AI Services Container, which allows organisations to run private AI models, such as embedding models or open-weight LLMs, within their own cloud tenancy or on-premises infrastructure, reducing exposure to third-party AI providers.

Oracle AI Database 26ai APIs now support integration with NVIDIA NeMo Retriever and NIM microservices, facilitating vector embedding and retrieval-augmented generation pipelines. The platform is designed to support future use of NVIDIA's GPU-accelerated vector search capabilities for further performance improvements.

For application development, Oracle introduces features such as data annotations to inform AI on data purpose and semantics, a unified data model that supports relational, JSON, and graph data, and new frameworks for building and deploying in-database AI agents with security and scalability in mind.

The update also introduces mission-critical innovations such as Oracle Database Zero Data Loss Cloud Protect, which guards against data loss and ransomware, globally distributed database capabilities for scalability and data sovereignty, a transparent middle-tier cache called True Cache, and a scalable in-database SQL Firewall for increased security against unauthorised or malicious database activity.