Status of this policy
SparkBench is in development. This policy describes what this website does today and what the add-in and service are designed to do. Where a fact is not yet decided, the policy says so instead of guessing. The sections on the add-in describe design commitments, not a released product, and will be revised against the actual implementation before any release or store submission.
Who is responsible
Pending. The legal identity and registration number of the organisation responsible for SparkBench and this website have not yet been published on this preview. They will be stated here before this policy is relied on for any product, store submission or commercial use.
Pending. A monitored contact address for privacy questions and requests has not yet been published. It will be added here and on the support page.
This website
What we set up
- Static pages only. There are no accounts, forms, comment fields, uploads, newsletters or embedded third-party content.
- No cookies are set by the site and no analytics or tracking scripts are included. There is nothing to consent to, so there is no consent banner.
- All scripts, styles, fonts and images are served from this site's own address. Pages make no requests to other services.
- The interactive demonstration runs entirely in your browser on invented data and sends nothing.
What the hosting provider processes
The site is served by Microsoft Azure Static Web Apps (Free plan; the resource is located in East US 2, with global static content distribution). To deliver a page, the provider's servers necessarily handle the ordinary technical data of a web request, such as your IP address, the address requested, the time and your browser's identification string. Public pages may be cached and served from locations around the world. Hosting location says nothing about where any future AI processing happens; that is covered below.
Pending. What the site's source and configuration establish is the absence of any application-level collection: no forms, accounts, analytics, tracking or cookies set by the site. They do not establish how the hosting provider logs requests, for how long, who at the publisher can access such logs, or the respective roles of the publisher and the provider for that data. Those facts have not been verified for this preview and are not asserted here. They will be stated when confirmed by the people who administer the hosting.
Purpose and basis
The purpose of the technical processing is delivering the pages you request and keeping the service secure. Pending. The legal basis under the General Data Protection Regulation, and whether the publisher or the hosting provider is responsible for which part of the processing, have not been determined for this preview and are not claimed here.
Support correspondence
No support address is published on this preview yet. When one is, this section will state what is processed when you write to it, for what purpose and for how long. The intended practice is: your address and message are used to answer you and are not used for anything else; retention period pending.
Retention: pending. A retention period for support correspondence has not been set.
The SparkBench add-in and service
Nothing in this section is live. It states the design that the add-in and its service are built and tested against, so that readers can evaluate it before a release. It will be reconciled with the actual implementation and completed with the pending facts before release.
Design rule
SparkBench acts as the remote AI's local representative with the user's authority. The remote AI receives only the material the user has reviewed and approved: the task description, the structure of the referenced data (table and column names and types), and shareable development material containing nothing from the user's workbook. Workbook contents, results and reports derived from them, faults and error messages, counts, later conversation context, and source code that captures private values remain on the user's device.
Categories of data, by destination
- Sent to the remote AI
Approved task text; data structure; shareable development material. The task text and column names are written or chosen by the user and may contain personal data if the user includes it; SparkBench shows them before sending and the documentation explains how to avoid that.
- Processed on the user's device only
Workbook contents, results, explanations, faults, counters and session history. Not transmitted to the publisher or the AI.
- Account, licensing and operational data
Pending. Whether the service will process user or tenant identifiers for licensing, and whether any operational metrics are collected, is undecided. The design rule excludes private-derived results, faults and counters from any telemetry.
Recipients
Pending. The AI provider and the account under which it is used are not yet selected. Two routes are under consideration: AI resources in the customer organisation's own cloud subscription, or AI hosted by the publisher. The provider's processing terms will be named here when the route is chosen. Public static hosting of this website or of the add-in's files does not determine where AI processing occurs.
Retention
Pending. Retention periods for any service-side data depend on the undecided items above and will be stated before release. Data processed on the user's device remains under the user's control in Excel.
What SparkBench does not change
A workbook stored in Microsoft 365 or opened in Excel on the web is already processed by Microsoft under the user's or organisation's agreement with Microsoft. SparkBench does not alter that, and this policy does not claim that all information stays on one device.
Your rights
Where the General Data Protection Regulation or similar law applies, you have rights to access the personal data held about you, to have it corrected or erased, to restrict or object to its processing, to receive it in a portable form where applicable, and to complain to a supervisory authority. This website itself collects nothing from you through forms, accounts or scripts; whether the publisher can access hosting-level request logs is pending, as stated above.
The contact for exercising these rights is pending, as stated above.
Changes
This policy will change as decisions are made. Each version carries a date at the top. Material changes to what the add-in sends or retains will be described in the version that introduces them, before they take effect.
What this draft does not yet settle
- The responsible organisation's legal identity and registration number.
- Monitored contact addresses for support and privacy requests.
- The AI provider, the account route and the applicable processing terms.
- Whether licensing identifiers or operational metrics are processed by the service.
- Retention periods for support correspondence and any service-side data.
- How the hosting provider logs requests for this website, who can access those logs, the legal basis for that technical processing, and the roles of publisher and provider.
- Languages in which this policy will be offered.
Until these are settled, this policy supports an honest description of a product in development. It is not presented as complete for commercial use.