Data controls for AI

Transform data before AI receives it.

Source access determines what Onterix may retrieve. At each handoff, release controls apply your policy to what a person, agent, model, tool, or destination actually receives.

See the release decision

See exactly what crosses the boundary.

Onterix checks current source access, applies the required built-in, DLP, and custom transformations, and records the exact result released to the recipient.

Data retrieved

5 fields
Account
Acme Manufacturing
Treatment
keep
Renewal
Sep 30 · $1.2M
Treatment
keep
Contact email
[email protected]
Treatment
tokenize
Internal risk note
Executive escalation watch
Treatment
remove
Support severity
High
Treatment
keep
ONTERIX RUNTIMEData release control

Release check

Before release
Requester
Maya Chen · person
Recipient
Maya Chen · Codex session
Purpose
Renewal brief
Source access
Current
Custom transform
v12
Destination rules
v8
Release decisionTransform1 removed · 1 tokenized

Data released

4 fields
Account
Acme Manufacturing
Treatment
release
Renewal
Sep 30 · $1.2M
Treatment
release
Contact
contact_7A2F
Treatment
release
Support severity
High
Treatment
release
Release evidenceSource revision sf_842Transform tx_12Output hash 7d9a41c2otr_01JQ9F42

One release decision can combine built-in checks, your approved DLP or classifier, and a customer-controlled transformation. If a required check fails, no result is released.

See how release controls fit the security model

Check every handoff

Recheck data when its purpose or destination changes.

The same data may be safe internally, barred from an external model, and safe again after transformation.

INGRESS

Prompt or upload

Check text and files before they enter AI context.

RETRIEVAL

Retrieved data

Inspect and transform source data before release.

ACTION

Tool input

Check proposed arguments and sensitive fields before execution.

DELIVERY

File or message

Recheck content, recipients, and attachments before delivery.

Connected AI clients and agents use the Onterix gateway. Endpoint inventories local AI paths and routes supported clients through the same controls, while uncovered paths remain clearly labeled.

Explore endpoint coverage

Control data release

Give each recipient only the data the job requires.

Start with built-in protection, add your DLP, or write a custom transformation. Each run records the release decision and exact data released.