Skip to main content

Security & trust

Trust is the product.

OrbitaAI handles information people would rather keep private. That shapes every architectural and product decision — including the ones that make us slower.

Our commitments

Four promises we can be held to.

These are product requirements, not aspirations. If we break one, it's a bug — not a policy debate.

01

You own your data

Export a portable copy or delete anything at any time — without a support ticket and without losing access to privacy controls.

02

Consent before action

OrbitaAI prepares work. It never sends, submits, pays, or deletes on your behalf without precise, explicit approval.

03

Evidence over confidence

Important facts link to their source. Uncertainty is labeled. The system does not invent deadlines, amounts, or quotes.

04

No data business

No advertising, no data brokerage, no training on your private documents. Revenue comes from subscriptions and licensing.

How it's built

Security decisions, in plain language.

We won't hide behind jargon. Here's what we do, what it means, and where the limits are.

Important distinction

OrbitaAI is not end-to-end encrypted. Understanding a document requires server-side processing, which means the service can access content while processing it. We limit that access, but we won't misrepresent the model.

Data minimization

A reminder doesn't need your inbox. A summary doesn't need your location. We request only the permissions a specific feature requires, and we avoid broad integrations in early access.

Encryption in transit

Traffic between your browser and OrbitaAI uses TLS. Data at rest is encrypted by our infrastructure providers. Access is limited to what's required to operate the service.

Least-privilege access

Team access to production data is restricted, logged, and reviewed. Support staff cannot browse your documents without a documented, user-authorized reason.

Untrusted input by default

Documents, OCR output, and messages are treated as data — never as instructions. Prompt-injection defenses and schema validation sit between the model and any action.

Model providers under contract

AI processing is performed by vetted providers under terms that prohibit training on your content. We don't use your documents to improve third-party models.

Retention with an exit

Deleting a document removes the item and the facts extracted from it. Backups age out on a defined schedule. Export is available before you delete.

Hard limits

What we will never do.

These aren't temporary policies. They're boundaries that define whether the product deserves to exist.

  • Sell your data or use it for advertising

  • Train AI models on your private documents

  • Monitor your devices or read your inbox covertly

  • Impersonate you or bypass another service's security

  • Take a consequential action without explicit approval

  • Claim a certification or audit we haven't completed

Your controls

Control where you'll actually look for it.

Privacy settings belong next to the thing they affect — not buried three menus deep.

Export

Where
Settings → Your data
What it does
Download a portable copy of saved information.

Delete

Where
Next to each item
What it does
Permanently removes the item and extracted facts.

Memory

Where
Memory manager
What it does
Review, edit, pause, or forget saved preferences.

Sharing

Where
Per item
What it does
Grant and revoke access without exposing the underlying document.

Notifications

Where
Reminder settings
What it does
Quiet hours, digest mode, snooze, and pause.

Incident response

If something goes wrong, we tell you.

If we become aware of a security incident affecting your personal data, we will notify affected users without undue delay, explain what happened in plain language, and describe the steps we're taking. We won't quietly bury it in a status page.

We do not currently claim SOC 2, ISO 27001, or HIPAA certification. As the product matures, we'll pursue independent audits when they provide meaningful assurance — and we'll publish the results.

Report a vulnerability

Found something that looks wrong? We want to hear it. Please include steps to reproduce and avoid accessing other people's data.

[email protected]

We'll acknowledge reports within 2 business days and keep you updated until resolution.

Early access

Ready to put your life in motion?

Join the early access list. We'll invite small groups as the product is ready — starting with document-to-action.

  • Free during early access
  • No credit card required
  • Your data stays yours

Get the first invite

One short email when your invite is ready. Nothing else.

Early access only. No spam, no credit card. Unsubscribe anytime. Privacy Policy.