Home/Security

Your AI only sees what each person is allowed to see.

No copied data, no new permission model, no credentials handed to an assistant. Optialy inherits the controls your systems already enforce.

Access logExample
ask Jordan via Copilot
check Jordan → Meta Ads read-only ✓
check Jordan → Shopify read-only ✓
check Jordan → Payroll no access ✕
rule Payroll ruled out, not guessed
keys credentials stay server-side
model bounded payload sent
answer cited, as Jordan
Access, step by step

Every request runs as the person asking.

Asked in Copilot byJordan, Marketing
Optialy checks accessRuns as Jordan
Meta Adsread-only
Shopifyread-only
Payrollno access, ruled out
Read-only by default

Any write needs explicit approval.

Credentials never reach the AI

Kept in managed secret storage, encrypted at rest.

Read in place

No warehouse copy. Caches are short-lived.

Bounded model payloads

Zero-retention model terms supported.

Revoke anytime

Access ends the moment a login is disconnected.

Independently tested

External penetration test at least yearly.

Trust by design

No AI writes a number the user acts on.

Our beta users told us a correct answer is not enough. They need to see where each number came from, how current it is and how it was calculated. We rebuilt around that.

01Does our data leave our systems?

No copy is made. Optialy reads what a question needs, when it is asked.

02Can people see data they shouldn't?

No. Everyone sees only what their own login already allows.

03Can the AI change anything?

Read-only by default. Any action needs a person to approve it first.

04How do we know a number is right?

Code computes every figure and each one links to its source. Unclear fields get a question, not a guess.

More questions security teams ask

The short answers.

Does Optialy copy our data into a warehouse?

No. Optialy reads each system where the data lives. Query results are cached for a bounded time only, and caches are kept separate from access decisions.

Can the AI assistant see our credentials?

No. Credentials are held in managed secret storage on our side and never passed to Copilot, Claude, ChatGPT or any other assistant.

Can Optialy change data in our systems?

Optialy is read-only by default. Writes are explicit, tied to a specific workflow, and run only after a person confirms.

Whose permissions apply?

The asking person's. Optialy checks the user's access and the connected source's access before use, so an assistant never reaches data that person couldn't.

What do model providers receive?

Bounded payloads, not your full tables. Zero-retention model-provider terms are supported.

Can we cut off access?

Yes. Disconnect a source and its access is revoked.

Is Optialy independently tested?

An external firm runs a penetration test at least annually, and vulnerabilities are fixed within set windows by severity.

Can we run Optialy privately?

Private deployment for regulated teams is designed and on our roadmap. Talk to us about your network and data-residency requirements.

Bring your security team in early. We will.

Start with one question that costs you money today. We'll answer it on your live data.