Oracle Deep Trust AI Brings Advanced Security and Governance to Enterprise AI Agents

Artificial intelligence is rapidly changing how businesses operate. Modern AI systems can generate software, analyze corporate information, automate workflows, and perform tasks that previously required human involvement. However, as companies increasingly depend on autonomous AI agents, protecting sensitive information and maintaining control over automated decisions have become major challenges.

To address these concerns, Oracle has introduced its Deep Trust AI Architecture, a framework designed to integrate security, privacy, and business governance directly into the underlying technology infrastructure.

Announced at Oracle AI World, the new architecture aims to help organizations deploy AI applications and autonomous agents with greater confidence. Instead of depending entirely on individual AI models or developers to implement security correctly, Oracle wants essential protections to operate directly within the database, infrastructure, data access systems, and application development environment.

Why Enterprise AI Needs Stronger Security

AI agents are becoming increasingly capable of performing complex business operations. They can access corporate databases, generate applications, process transactions, and make decisions based on organizational information.

However, these capabilities also introduce new security risks.

An AI agent could accidentally expose confidential information, misunderstand a business instruction, generate an insecure application, or modify important records incorrectly. Traditional security measures that depend heavily on application-level programming may not always provide sufficient protection.

Oracle believes that organizations need a more comprehensive approach in which security and governance are built into the systems that AI agents use.

The company points to Gartner's projection that 40% of enterprise applications will include task-specific AI agents by the end of 2026, compared with less than 5% in 2025. This highlights the growing importance of reliable AI infrastructure as businesses expand automation.

Oracle Deep Trust AI Architecture Uses Four Security Layers

Oracle's new architecture combines several technologies into four interconnected layers. Each layer addresses a different aspect of enterprise AI security and helps organizations maintain control over automated systems.

1. Software Platform: Protecting Enterprise Infrastructure

The first layer focuses on securing the infrastructure that supports enterprise applications and databases.

Oracle Autonomous AI Database provides automated security patching across the database engine, operating system, Exadata storage, and firmware. These updates help organizations address vulnerabilities while minimizing disruption to running applications.

The platform also supports rolling updates, allowing maintenance activities to take place without requiring conventional application downtime.

Another important feature is always-on encryption, which protects information both when stored and while being transmitted.

By strengthening the underlying infrastructure, Oracle aims to reduce security risks before they affect applications and AI-powered services.

2. Data Layer: Enforcing Privacy and Business Rules

The second layer focuses on protecting enterprise information and controlling how it can be accessed or modified.

Oracle Deep Data Security allows organizations to establish access policies directly within the database. These controls use cryptographically verified user identities to determine which information an individual or AI-powered application is authorized to access.

This approach reduces dependence on application developers to implement every security requirement correctly.

Oracle is also introducing SQL Assertions, which allow administrators to establish business rules that database transactions must follow.

For example, an organization could prevent a transaction from exceeding an employee's authorized spending limit. If an AI agent or application attempts an update that violates the rule, the database can reject the transaction.

This provides an additional safeguard against unauthorized or incorrect changes, regardless of which application or agent initiates them.

3. Data Access Layer: Controlling How AI Agents Use Information

Giving AI agents unrestricted access to corporate databases can create serious security concerns.

Oracle addresses this challenge through Deep Trust Data APIs, which provide controlled access to business objects such as purchase orders instead of allowing agents to interact freely with underlying database tables.

This makes it easier for organizations to restrict AI operations to the information and actions required for a particular task.

Another important component is Oracle Deep NL2SQL, which helps translate natural-language questions into database queries.

The technology uses business and data context to improve how AI understands organizational terminology, metrics, and information structures.

Oracle also incorporates Continuous Feedback Learning, allowing users to correct inaccurate answers, alongside Knowledge Change Verification to validate corrections before incorporating them into the organization's knowledge base.

These capabilities aim to improve the reliability of AI-generated business insights and reduce misunderstandings when employees interact with corporate data through natural language.

4. Application Development: Building More Secure AI-Generated Software

The fourth layer focuses on improving how AI-generated applications are developed and executed.

Oracle APEX AI Application Generator allows AI coding agents to transform natural-language application descriptions into APEXlang, an open, declarative application definition language.

Instead of generating large amounts of conventional application code that organizations must independently secure and maintain, AI agents produce application definitions that are executed by the Oracle APEX runtime.

The runtime provides essential enterprise capabilities, including security, scalability, and availability.

Developers can use familiar tools such as Visual Studio Code, Oracle APEX, and AI coding agents including Codex.

This separation between AI-generated application definitions and the underlying execution environment is intended to make enterprise application development more manageable and secure.

Oracle Enters the Growing AI Governance Competition

Oracle is entering an increasingly competitive market for enterprise AI governance.

Companies such as Snowflake and Databricks already provide technologies that help organizations manage data access, governance, and AI-powered workloads.

Oracle's main advantage is its ability to integrate security controls directly into its database infrastructure, Exadata platform, and APEX application runtime.

This approach is particularly relevant for organizations that depend on Oracle databases to manage important business transactions.

For example, database-level restrictions can prevent unauthorized changes even when those requests originate from newly developed AI agents or applications.

Oracle is also expanding its capabilities beyond traditional relational databases through Autonomous AI Lakehouse. The platform supports enterprise data management across Oracle and non-Oracle environments, including open Apache Iceberg tables.

Its Business Semantic Layer is designed to help organizations maintain consistent business definitions and improve how AI-powered systems interpret enterprise information.

These developments strengthen Oracle's position as businesses look for ways to manage AI workloads across different cloud and data environments.

Why Deep Trust AI Could Matter for Businesses

The biggest challenge facing enterprise AI is no longer simply building more intelligent models. Organizations also need to ensure that AI systems operate within clearly defined security boundaries.

Oracle's approach focuses on making important restrictions part of the underlying infrastructure rather than relying exclusively on AI-generated code or application-level safeguards.

For businesses already operating critical workloads on Oracle databases, this could simplify the process of introducing autonomous agents into production environments.

Security teams can establish policies that apply consistently across applications, employees, and AI agents. Database-level business rules can also help prevent unauthorized transactions and incorrect updates.

However, organizations will still need to evaluate implementation requirements, configuration, access policies, and operational risks before allowing AI agents to perform sensitive business activities.

The Future of Trusted Enterprise AI

Oracle Deep Trust AI represents a broader shift in how businesses approach artificial intelligence.

As AI agents become more involved in corporate operations, companies will need more than powerful language models and automated workflows. They will require dependable infrastructure capable of protecting sensitive information, enforcing business policies, and maintaining control over automated decisions.

By combining infrastructure security, database-level governance, controlled data access, and trusted application development, Oracle aims to provide a unified foundation for enterprise AI adoption.

The real test will be how effectively organizations can use these capabilities in everyday business operations.

If Oracle successfully delivers on this approach, Deep Trust AI could become an important part of the company's strategy to help enterprises move autonomous AI systems from experimental projects into business-critical production environments.