Oracle today announced new agent AI innovations for Oracle AI Database. It enables customers to quickly build, deploy, and scale secure agent AI applications suitable for serious production workloads. Oracle AI Database integrates and architects agent AI and data across operational databases and analytics lakehouses. This enables AI agents to securely access real-time enterprise data wherever it resides and easily use LLMs trained on public data and business data to provide business insights. Customers can choose their AI models, agent frameworks, open data formats, and deployment platforms. Additionally, customers running Oracle Exadata will further benefit from Exadata Powered AI Search. This enables highest-scale agent AI with faster AI queries for high-volume, multi-step agent workloads.
“The next wave of enterprise AI will be determined by the ability of our customers to use AI in their business-critical production systems to securely deliver breakthrough innovation, insight, and productivity,” said Juan Loaiza, executive vice president of Oracle Database Technologies at Oracle. “With Oracle AI Database, customers don’t just store data, they activate it for AI. By architecting AI and data together, we help customers quickly build and manage agent AI applications that can securely query and interact with real enterprise data with stock exchange-grade robustness on all major clouds and on-premises.”
Accelerate innovation with AI designed for data
With agent AI capabilities designed for data, Oracle AI Database eliminates the need to build and maintain data movement pipelines that increase complexity, security risks, and can produce worse outcomes. New features include:
- Oracle Autonomous AI Vector Database Delivering the simplicity of a vector database with the full power of Oracle AI Database. This allows developers and data scientists to quickly and easily build vector-powered applications using an intuitive API and easy-to-use web interface. Built on the Oracle Autonomous AI Database, it combines an easy-to-use developer experience with enterprise-grade security, reliability, and scalability. Currently with limited availability, the Autonomous AI Vector Database is accessible through the free tier or low-cost developer tier of Oracle Cloud. As your requirements grow, customers can seamlessly upgrade the full functionality of Oracle Autonomous AI Database with one click, with full support for graph, spatial, JSON, relational, text, and parallel SQL, eliminating the need for separate databases or complex cross-database agent workflows.
- Oracle AI Database Private Agent Factory Business analysts and domain experts can quickly build and securely deploy data-driven agents and workflows. AI Database Private Agent Factory provides a code-free AI agent builder that runs as a container in the public cloud or on-premises, maintaining data security by enabling customers to build, deploy, and manage AI agents without sharing data with third parties. AI Database Private Agent Factory includes multiple pre-built data-focused AI agents, including a database knowledge agent, structured data analytics agent, and deep data research agent. Other approaches rely on external agent orchestration or require calls to different types of databases. Oracle has simplified agent AI for business users by building it into an AI database, delivering consistency and simplicity with enterprise-grade security, resiliency, and scalability for any agent workload.
- Oracle Unified Memory Core Users can store AI agent context in a single system. This uniquely enables low-latency inference across vector, JSON, graph, relational, textual, spatial, and columnar data in one integrated engine with consistent transactions and security.
Minimize AI data risk
Oracle AI Database helps protect data from external attacks, insider abuse, accidental disclosure, and unintentional exposure to LLM across multicloud, hybrid, and on-premises environments. New features include:
- Oracle Deep Data Security Implement powerful end-user-specific data access rules in your database. Each end user, or an AI agent acting on behalf of an end user, can only see the data that the end user is allowed to see. You can implement advanced persona and feature-based rules. For example, specify which parts of a customer account can be seen by specific sales representatives, finance personnel, shipping personnel, executives, support personnel, and customers’ relatives. This provides unique end-user data security capabilities that protect against new AI-era threats such as prompt injection using declarative, database-native controls that implement least privilege access. By centralizing and separating security from application code, customers can easily determine who can see what data and continually update access rules as new threats emerge, effectively providing guardrails for agents working within the Oracle AI Database. The security of the database, the source of the data, provides excellent protection when an AI agent directly accesses the data on behalf of the end user.
- Oracle Private AI Service Container This allows customers with stringent security requirements to run private instances of their AI models while avoiding data sharing with third-party AI providers or sending data outside their firewalls. Additionally, customers can safely offload compute-intensive AI tasks such as vector embedding generation outside of the database, reducing performance bottlenecks and keeping all data in the environment safe. Containers can be deployed in public clouds, private clouds, or on-premises, including air-gapped environments.
- Oracle Trusted Answer Search It provides businesses with an accurate, testable, and definitive way to use AI to provide answers to end users. Rather than using LLM directly to answer end-user questions, Trusted Answer Search uses AI vector search to match questions to previously created reports. This helps reduce the risk that the probabilistic LLM may hallucinate or misinterpret the query.
Eliminate AI data lock-in with open standards and frameworks
Oracle AI Database runs on all major cloud providers, hybrid deployments, and is available on-premises, giving customers the flexibility to choose the AI model and application-layer agent framework that best fits their needs. Build, deploy, and run agent AI applications using open standards and data formats. New features include:
- oracle vector on ice provides customers with native support for vector data stored in Apache Iceberg tables. AI Vector Search can read vector data directly from Iceberg tables, create vector indexes to speed up vector searches, and automatically update these indexes as the underlying vector data changes. Oracle Vectors on Ice enables AI search against data lake data, enabling unified search across business data in the database and vectors stored in the data lake. This enables customers to achieve unified intelligence across databases and data lakes.
- Oracle Autonomous AI Database MCP Server Enables external AI agents and MCP clients to securely access Autonomous AI Database and its features without custom integration code or manual security controls. It complements the Oracle SQLcl MCP Server for Oracle AI Database, available through the Oracle SQL Developer VS Code extension.
said Steven Dickens, CEO and Principal Analyst at HyperFRAME Research. “In the age of agent AI, a unified memory core is essential for agents to maintain context across a variety of data types, including vector, JSON, graph, columnar, spatial, text, and relational, without the delays and staleness caused by external synchronization.” “Oracle AI enables this in a single mission-critical engine with concurrent transactional and analytical processing, high availability, and robust security. “Organizations without this foundation will struggle with fragmented and unreliable agents, while organizations leveraging Oracle will gain a decisive advantage in scalable AI deployments.”
Customers and developers can take advantage of new agent AI capabilities in Oracle AI Database to begin developing and deploying innovative agent AI applications without moving data, learning new skills, or struggling with database scalability or lack of agent AI security guard rails. To learn more about the latest AI innovations, check out this Oracle AI Database Agentic AI announcement blog.
