What Happens to a Photograph You Upload

Uploading a photograph to an online service means handing an image of your face to a third party. That is an ordinary operation, but it helps to know what happens to it and which questions are worth asking beforehand.

The typical path of an image

1. Upload: the file is transferred to a server, usually over a secure connection.
2. Validation: checks on format, size, sometimes content against the rules.
3. Processing: the image is passed to a model that produces the result.
4. Storage: source and result are kept for a period — to display to you, for re-download, for handling support requests.
5. Deletion: after a set period or on your request.

Questions worth finding answers to

• How long are the uploaded image and the result stored?
• Are uploaded images used to train models?
• Are they passed to third parties — a model provider, for instance?
• Can you delete your data, and how quickly does that happen?
• Where is the data physically stored?

The answers are usually in the privacy policy and terms of service. Differences between services are substantial, and assuming they are the same is wrong.

Training on user data

A separate and the most sensitive question. Some services state explicitly that they do not use uploaded images for training; some reserve the possibility; some say nothing.

The practical significance: an image that entered a training set influences the model and cannot be "withdrawn" by deleting a file. So this point is worth checking before uploading rather than after.

Transfer to providers

Many services do not run models themselves but call external providers. That means the image passes through another organisation with its own terms.

It is practical to check whether this is stated in the policy: its presence is a sign of transparency rather than a problem in itself.

What not to upload

• Photographs of other people without their consent.
• Images of children.
• Documents and anything containing personal data in frame.
• Photographs revealing private details: an address, a routine, a workplace.
• Material you would not be comfortable seeing outside the service.

Upload hygiene

1. Crop before uploading: surplus in the frame is surplus data.
2. Check the background for private details.
3. Strip geotags and metadata from the file if the service does not do it.
4. Delete results and sources from the service once they are no longer needed.
5. Do not upload the same file repeatedly — every upload is a new copy.

What to do when unsure

A practical order: read the policy sections on storage and training, check that deletion is available, start with one unimportant photograph, and only then work with material that matters.

And a general point: the more sensitive the image, the higher the transparency you should require. For an avatar from a holiday snapshot a basic check suffices; for a family archive it is worth looking more carefully.