Oracle announced its Deep Trust AI Architecture on Oct. 26 at Oracle AI World in Las Vegas, proposing database-level controls for enterprises connecting AI agents to business information and operational actions.
The company describes a blueprint spanning the software platform, data, data access and application development. It combines Oracle AI Database, Autonomous AI Database and APEX AI Application Generator.
Moving rules closer to data
According to Oracle’s announcement, Deep Data Security applies end-user access policies within the database, while SQL Assertions enforce business rules. Deep Trust Data APIs restrict agents to defined business objects, such as purchase orders, rather than unrestricted SQL access.
The purchasing implication is a change in where controls reside: a buyer evaluating an agent also needs to examine the database permissions and interfaces through which it acts. These are Oracle’s descriptions of its architecture, not independently verified guarantees against misuse.
A framework is not a deployment result
For broader context, the National Institute of Standards and Technology describes its AI Risk Management Framework as voluntary guidance for incorporating trustworthiness into AI design, development, use and evaluation. NIST says version 1.0 is being revised.
That guidance is separate from Oracle’s product announcement and does not certify this architecture. For an enterprise evaluation, the unresolved practical question is how the proposed controls work with the organization’s actual identities, data and workflows—not whether a product name promises trust.