At Oracle AI World in Las Vegas, Oracle announced the Deep Trust AI Architecture. Think of it as Maximum Availability Architecture (MAA) for AI trust. The guardrails for agents and AI-built apps live in the stack, not in each application.
The four layers
- Software platform: Autonomous AI Database applies security updates automatically, and rolling patching keeps applications running. Encryption is always on.
- Data: Deep Data Security enforces each user's access rules on every SQL request. SQL Assertions reject any change that breaks a rule you've declared, whichever application, user or agent makes it.
- Data access: agents read and write whole business objects, such as purchase orders, through Deep Trust Data APIs instead of arbitrary SQL. Deep NL2SQL adds business knowledge to natural-language queries and re-verifies previously validated "gold" answers before a knowledge change is published.
- Application development: the APEX AI Application Generator turns natural-language app descriptions into APEXlang, and APEX runs those definitions in its runtime engine.
What you can do today
Oracle gave no availability dates for the newer pieces, so check before you plan on them. Several parts already ship. APEXlang is in APEX 26.1 and later, and annotations and SQL Assertions are in Oracle AI Database 26ai. Annotate your key tables, move one cross-table rule into an assertion, and keep agents away from raw tables.
My take: trust moves out of the prompt and into the database. Sources: Oracle press release and Announcing the Oracle Deep Trust AI Architecture.
Get the Oracle + AI digest
One email when a new tested guide, template or APEX release note lands — English or Arabic. No spam; unsubscribe anytime.