Privacy FAQ

Thank you for trusting Mathpix with your images! We take this responsibility very seriously. Below you can find easy to understand answers to common questions we get about our privacy policy and how we handle your (or your end-user’s) images and data.

You can also read the full Mathpix Privacy Policy.

What happens to our end-user’s images when using Mathpix OCR API in our product?

When using Mathpix OCR API to process images of math and text, each image or batch of images is sent to our servers where we either return a valid response or an error message if we cannot process the image for any reason. (More information on API request options, formats, and errors can be found on our Docs).

Unless a customer has specifically requested that all images sent from their API key be deleted from our database immediately after processing, once the image has been processed the image is then sent to our internal Quality Assurance (QA) database. These images can be accessed by our QA team, in order to assure that our service is working as intended and for finding any bugs in our recognition engine.

Any images with personally identifying information are deleted from the QA database within 24 hours. All images are deleted from the QA database after 90 days.

What if we want Mathpix to delete all images processed from our API key immediately?

There are two ways to indicate to Mathpix that you never want us to save any images to disk:

Per API key. Open the key’s settings from the API Keys tab on console.mathpix.com and turn Improve Mathpix off. See Managing your keys.
Per request. Include the improve_mathpix flag in your request, set to false.

The key setting takes precedence over the request flag, in both directions. Once Improve Mathpix has been set on a key — on or off — that decision governs every request made with it, and the per-request flag has no effect. The flag only decides the outcome for keys where the setting has never been touched, and in that case retention is on unless the request explicitly sets improve_mathpix to false.

The practical consequence: if you need a guarantee that nothing from a key is retained, set it on the key. Do not rely on every call site remembering the flag, and do not assume a false in the request will override a key that has the setting turned on.

Please note that querying v3/ocr-results — and browsing the Console Results tab — will only work for requests that were retained. Turning retention off is a deliberate trade: nothing kept, and nothing to look up afterwards.

How long are uploaded documents and page images retained?

The QA database retention described above (90 days, controlled by improve_mathpix) is separate from the retention of v3/pdf processing artifacts. Uploaded source documents and the page image files that back cdn.mathpix.com/cropped/... URLs are retained for up to 30 days. Mathpix Markdown and JSON Lines outputs are retained for up to 90 days. Setting improve_mathpix: false deletes the uploaded source document immediately after processing but does not change retention of page images or text outputs.

If you need image URLs to remain accessible beyond 30 days, request a zip output format (.mmd.zip, .docx, .html.zip, .tex.zip, etc.) at processing time — zip outputs embed all images inline. See the Data Retention page for the full retention matrix, long-term storage patterns, and how to request extended retention on your account.

Does Mathpix collect any information about my end-user?

Mathpix collects IP addresses of end-users to protect us in the case of spamming of our servers. Not collecting end-user IP addresses is an available option, but please note that if spamming is coming from your API key and we do not have IP information to locate the source, we will be forced to suspend your API key until the issue has been resolved.

What does Mathpix do to keep images safe?

We take our data management extremely seriously, and we have taken many measures to ensure that all data is managed in a way that protects your privacy. We only require the minimum amount of personal information that is necessary to fulfill the purpose of your use of Mathpix OCR API in your product.

Mathpix uses third party vendors and hosting partners for the necessary hardware, software, networking, storage, and related technology required to provide our services. Although Mathpix owns the code, databases, and all rights to the API code, you retain all rights to any data sent from your API key.

In order to provide the best quality of service, we must monitor the images being processed by the app. Any image reviewed by our QA team has already been completely anonymized from your end-user. Only those responsible for monitoring server health have access to end-user IP addresses. All images are deleted from our QA database after being reviewed by our QA team (within 90 days).

Please note that our policies and practices are subject to change–we are always trying to improve how we handle the data we process. Any changes to our policies will be reflected on our website in our Privacy Policy, Terms and Conditions or in this FAQ.

How can I request access to the data Mathpix has from my API key?

You have access to all the API request and response data from your API keys under the Results tab of your OCR API Organization page. See Results for what the tab shows and how to query the same data from your own code.