# Privacy and security of your label data

> Your label and company details remain private during AI-assisted food label analysis. We explain processing purposes, deletion of original files and access to retained reports.

Source: https://gidaetiketi.com/en/privacy

## Who operates the service?

Gıda Etiketi is a preliminary label checking service provided by Brix Mühendislik. Send data processing or privacy enquiries to ufuk@brix.tr. Our [contact page](/en/contact) provides contact details and the [Brix background page](/en/background) explains the engineering approach behind the service.

## Are labels and company details shared?

Your label data, product details and company information are not publicly published, sold or shared with third parties for advertising, marketing or another company’s commercial use. Uploads are not turned into public examples, public reports or advertising content.

This commitment does not exclude the technical processing by infrastructure providers described below, report emails you request or legally mandatory requests from competent authorities. It is not a claim that no third party ever processes the data.

## Uploaded files and extracted information

The selected label file, its name and size are processed for the check. Analysis produces extracted text, field evidence, rule results and job status. Source files are stored privately and are not published in a public file directory.

Do not include personal or confidential documents that are unnecessary for the label check. Review every page of a multi-page PDF before submitting it.

## Analysis access and administrator review

The access key created after upload is stored in the current tab’s browser session. Checking job status, viewing the results and downloading the report require that key. An analysis number is not a substitute for it.

Authorized Brix administrators may review original PDFs/images, retained reports, extracted text, evidence and feedback to investigate errors. Once the original PDF/image is deleted, administrators cannot open that file from the archive either. Rule approvals and expert reviews remain traceable as separate records.

## Server infrastructure and visual assessment

Analysis does not run only on your device. The uploaded file is processed in the service’s Cloudflare server and private storage infrastructure; OCR extracts text and field evidence. When visual assessment is enabled, label images may also be processed by Cloudflare Workers AI to extract evidence such as text emphasis or symbols.

If you request an emailed report, the recipient address and PDF are processed by the email delivery service. Processing in cloud services does not mean a file is publicly published. Do not upload personal or commercial documents unnecessary for the check. The [file security and privacy answers](/en/faq#security) explain this workflow in practical terms.

## Retention of original label files

Your original PDF or image remains in private storage. Completing or cancelling an analysis, closing the tab or the passage of time does not automatically delete it. Only authorized administrators can retrieve originals; no public file link is created.

The original remains until an administrator explicitly confirms its deletion. The action is recorded. Technical failures retry only that approved deletion request. Archiving does not delete anything, and expiry of the window for starting a new analysis does not delete the original.

Reports, extracted text, findings and expert reviews are retained separately and are not automatically deleted based on age. Deleting the original does not delete the report. Technical processing, application error, analytics and email event logs have a 90-day retention period. Use the contact channel below to request deletion of your data.

## System learning and accuracy improvements

Authorized staff may review label data to provide the service, investigate reading errors and improve this system’s OCR and rule checks. This does not permit external sharing of your label or company details.

Uploads are not automatically added to AI model training, a permanent training pool or the image template catalogue. Administrator-added templates are managed separately from user uploads. Feedback and expert reviews do not silently overwrite an existing machine result; proposed corrections are reviewed separately.

## Access protection

Files are stored privately. Report endpoints require an access key and administration endpoints require an administrator session; the server stores verification hashes of session and report keys. File format, size and request frequency are checked.

Reports and label contents are not included in search pages, sitemaps or public AI documentation. Share access keys and downloaded PDFs only with authorized people. These controls are not an independent security certification or a guarantee of zero risk.

## Visit analytics and technical error logs

We record UTM source/medium/campaign, landing page, referrer domain, device, browser, operating system, language and approximate country/region/city. Analytics does not store raw IPs, label content or full URL queries. A random tab session ID expires after 30 minutes of inactivity; no analytics cookies or device fingerprints are created. Administrators and detected bots are excluded. DNT/GPC disables visit analytics.

Analytics and technical error records are removed by daily cleanup after 90 days. Browser error types and code locations may be recorded to diagnose faults. Request bodies and form fields are not collected for error logging.

## Requested emails and optional notifications

Report emails are sent when requested through the report form. Delivery uses the recipient email address and report information. Delivery status and errors are recorded to track the request.

Emails are sent through Amazon Web Services SES. Content, recipient, subject, SMTP acceptance and verified delivery/bounce/complaint events are retained for 90 days. Permanent bounces or complaints prevent further sending to the address.

Consent for software updates and news about related software is not preselected and is not required to receive a report. Granting it creates a consent record. You can withdraw consent through the notification preferences link; this does not prevent later report emails that you specifically request.

## Feedback and service operation

The name, surname, profession, email and explanation submitted in feedback are used to investigate the finding. Avoid including unnecessary personal information. Processing logs help establish which analysis step completed or failed.

Request limits are used to reduce misuse of the service. Label contents are not included in public search pages or sitemap files. The English interface uses the same access controls and retention rules.

## Access, correction and deletion requests

Email ufuk@brix.tr with your analysis number and request to ask about processing, access, correction or deletion. Do not email your password or report access key. Ownership must be verified to prevent an unauthorized person from obtaining or deleting your report.

Requests to learn whether your personal data is processed, its purposes and recipients, correct inaccurate information or request deletion where applicable are reviewed through this channel. Email delivery and notification consent records, administrator change history and copies you downloaded are separate from the report archive; specify which records your request concerns.