Accenture Uses Oracle to Secure AI Agents in the Database
Accenture uses Oracle Deep Data Security to control AI agent access. This protects data even if agents write complex queries.
Key points
- Accenture uses Oracle Deep Data Security to enforce rules at the database level.
- The system manages access for over 40 agents across 925,000 documents.
- Tests showed 300 simultaneous queries returning results in under 15 seconds.
Accenture recently tested a new security approach for AI agents. They used Oracle’s Deep Data Security tool. This shifts trust to the database layer.
The system works with over 925,000 documents. Different agents have different access levels. The database enforces these rules on every query.
Moving trust to the database
Traditionally, apps control who sees data. But AI agents can act unpredictably. So Accenture moved the control point down to the database.
Oracle’s tool defines specific privacy rules for each user. The database applies these rules automatically. This happens regardless of the agent’s code.
This creates a hard boundary. It stops unauthorized access at the source. Users can trust that data stays protected.
Testing scale with many agents
Accenture built a multi-tenant application for this test. It supports many different users and agents. The system needs to handle high traffic.
They tested it with 40 or more agents. These agents hit the application simultaneously. The setup proved robust under load.
The system handled 300 simultaneous queries. Most results returned in 15 seconds or less. This shows the setup can scale effectively.
The trade-off for this method
This approach requires strict database configuration. It adds complexity to the backend setup. Teams must define rules carefully.
