Bind the human identity
OCI IAM validates the browser session. Safe claims and correlation identifiers remain server-side while raw credentials and tokens are never displayed.
Identity validated
OCI Responses · Oracle DeepSec · evidence by design
AI can reason.
AI can reason.
Oracle decides.
Meridian Guard lets OCI Responses select tools and propose SQL while Oracle DeepSec independently controls which payroll data may leave the database for the signed-in user.
OCI IAM Authenticated end user
OCI Responses Dynamic tools and SQL
Oracle DeepSec Database authorization
The control plane beneath the conversation
Freedom to reason. Authority stays with Oracle.
The application orchestrates OCI Responses and seven payroll tools. The model remains free to reason and propose SQL; Oracle DeepSec independently controls protected data release under the propagated end-user identity, with every boundary crossing retained as evidence.
Let the model use tools
OCI Responses receives the approved tool contracts on each request and can discover the data source, construct SQL, iterate, analyze numerical results and reconcile totals.
Seven tools attachedKeep authority in the database
Oracle DeepSec evaluates Data Grants under the propagated end-user context before protected payroll rows, columns and cells leave Oracle Database.
DeepSec governedEnterprise payroll intelligence
Ad-hoc questions without application-side authorization logic.
Payroll and workforce questions remain flexible, including totals, averages, percentiles, annualized values and reconciliation. Requests outside the payroll workspace receive a professional redirect; tax, legal, exchange-rate and other externally governed calculations are not invented without an approved source.
Assurance is part of the runtime