Oracle APEX Trusted GenAI Demo: Same Prompt, Different Authorized Data

This Oracle APEX 26.1 demo shows a trusted generative AI pattern inspired by Oracle Deep Data Security: identity first, database policy second, AI response last.

The app demonstrates a simple but powerful enterprise AI concept: the same prompt should return different authorized data depending on who is signed in. Alice, a sales representative, can only see her authorized accounts with sensitive fields restricted. Frank, a finance analyst, can see broader finance data because his role allows it.

In this local Oracle APEX build, Oracle Virtual Private Database enforces the database policy path. In a production Oracle AI Database 26ai deployment, the same trust model maps to native Oracle Deep Data Security end-user context, data roles, and data grants.

Fictional customer data is used for demonstration only.

Chapters

0:00 The AI data exposure problem
0:35 Role matrix and expected policy
0:54 Alice open-access baseline
1:16 Alice database policy path
1:38 Frank broader authorized access
2:00 Prompt injection does not expand data
2:16 Audit evidence and trust sequ

Видео Oracle APEX Trusted GenAI Demo: Same Prompt, Different Authorized Data канала AI Action Lab