Log in

Privacy policy

Last updated: 2026-10-02

The short version

We never ask you to upload your real architecture diagram or describe your real system in confidential detail. Matching works entirely against a library of generic, industry-standard reference architectures. The component names you type when confirming a diagram stay in your session and are never used to grow our library or train anything.

What we collect

  • Wizard answers. Your industry selection and the multiple-choice answers you give during matching (e.g. "inbound", "event-driven"). These describe the shape of an integration pattern, never your real system.
  • Match-miss feedback (optional). If you tell us what's different about your architecture when confidence is low, that free-text note is recorded to help us grow the reference library.
  • Component names. When you confirm a diagram, the names you type stay in your session (and, once you sign in, in your account) so your generated artifacts read in your vocabulary. These are never sent anywhere else and never used to grow our library.
  • Email address. Collected when you sign in to unlock checkpoints, test paths, plans, and reporting, or when you request the demo. Used to identify your saved workspace and to reach you about product updates — never sold, never spammed.
  • Progress and report data. Checkpoint pass/fail status you record, and any notes you add to a status report, are stored with your account so your workspace persists between visits.

What we never collect

In the architecture product we never collect real architecture diagrams, real system credentials, production data or payloads, or any confidential business information. Automation Studio works differently, because it tests a real application — see Automation Studio below. Sample test-data payloads generated in the product are synthetic and generic — editing them in your real expectations happens entirely in your browser and is not transmitted anywhere by that action alone.

How the plain-English summary works

The summary may be generated with the help of a third-party large language model, using only the generic pattern facts and the component names you provided — never your wizard answers' underlying business context beyond what you typed, and never used to train any model.

Test Utilities — a different data model

Everything above describes the core product: matching, checkpoints, paths, and reporting, none of which ever require your real system. Test Utilities is the one deliberate exception. Tools there — starting with the Payload Chain Builder — work directly with real sample JSON you provide: field names, structure, and example values from your own APIs or documents.

  • Kept private to your account. Uploaded payload schemas and field-dependency rules are stored only against your own login, never analyzed in aggregate, and never used to grow any shared library — the same treatment component rename strings get elsewhere in the product.
  • You control what you paste. We recommend using representative/fake values rather than real production data, same as the Test Data payloads generated elsewhere in the product — the tool only needs real field names and structure to work, not real data.
  • Deleted with your account. Right-to- erasure (below) removes saved chains along with everything else tied to your login.

Applications and Automation Studio — a real app

When you add an Application, you give us the URL of a web app you confirm you own or are authorized to test, and Automation Studio (in closed beta) runs a real browser against it. That means we handle information about that app, not a generic pattern.

  • What we store about the app. The pages the explorer reaches — headings, buttons, form fields and their declared rules, visible text, and the API paths each page calls — plus business rules quoted from those pages, the test scenarios, and each run's results, including a screenshot of every step and the requests it made. Screenshots show whatever your app displayed to the test account, so point it at a staging site or a test account, never real customer data. Old screenshots may be removed after a retention period; the results themselves are kept.
  • Test logins and other test data. Values you give it — such as a test account's password — are stored encrypted, are never shown back to anyone once saved, and are never sent to an AI model; they are only typed into your own site during a run. Each person can keep their own values, which nobody else can use or see. Use test accounts only, never a real person's login.
  • What AI models see. To write and repair scenarios, explain failures, read your pages' stated rules and compare your manual test cases, we send page structure and text, scenario text and the pasted test cases to Anthropic's Claude. When a step's target can't be found without it, the page's accessibility tree is sent to Claude as well, to locate it. No other AI provider receives your pages. Test data values are never included, and nothing is used to train any model on our side. If your workspace adds its own Anthropic key, these calls are made with that key, under your own agreement with Anthropic.
  • Credentials you connect. An Anthropic key, a Jira or Azure DevOps token, a single sign-on client secret and two-factor keys for test logins are stored encrypted and never shown back. Private-runner and deploy-hook tokens are stored only as a hash. When you file a failed run to Jira or Azure DevOps, the defect's title, steps and failure details are sent to that tracker — test data values are not.
  • Private runner. If you run an Application on your own private runner, the browser runs inside your network: step screenshots and traces stay on your machine, and only the results (step outcomes, the requests made, the explored pages) are sent back to us.
  • Private to your workspace. Everything above belongs to its workspace — your personal one, or a team you were invited to, whose members can see that team's Applications — and is never pooled across customers or used to grow our shared knowledge base. We record how many model tokens each workspace uses, not what was in the calls, and an activity log of who changed settings and credentials (key names only, never values).
  • Deleted with the Application. Deleting an Application deletes its pages, rules, scenarios, runs, screenshots and test data. Deleting your account deletes everything in your personal workspace; Applications you created in a team stay with that team, under one of its owners, and your own test data values are removed.

The blog

We count how many times each post is read, to show "Most read". The count is a number per post: nothing about you is stored with it. Your reading progress, points and player settings stay in your own browser and are never sent to us.

Deletion

You can clear your local session at any time with the "Start over" control in the header, and delete your account — with everything tied to it — from your Profile page. For anything else, email us at support@qualityaiq.com.

Contact

Questions about this policy? Email support@qualityaiq.com. QualityAIQ is a product in active development; this policy will be updated as the product matures, and material changes will be reflected on this page.