Some AI services describe using submitted material for model development, while others restrict that use for particular products or account types. The answer for your upload depends on the applicable terms, privacy notice, settings and law. It cannot be inferred from the word AI alone.
Read the documents for the exact product you use. A statement about a business service may not apply to its consumer version, and a setting available in one account may not exist in another.
Processing is not the same as training
Processing an input to produce a response is one activity. Using examples to adjust a model's parameters is another. A provider may also describe evaluation, moderation or troubleshooting. These purposes should not automatically be treated as synonyms.
The phrase “improve our services” is too broad to answer every question by itself. Look for an explanation of the data involved and the actual activities covered. Does the language refer to images, written instructions, feedback or usage statistics? Does it distinguish routine submissions from material deliberately sent to support?
Understanding what happens during an upload helps you ask a narrower question than whether a provider “uses your data”. Any functioning online service processes some information; the purpose and limits matter.
Check the scope of account controls
If a training control is available, read what it changes. Ask whether it applies to future submissions, previously stored material or only a particular workspace. Record when you changed it and retain the explanation shown alongside the setting.
Imagine a hypothetical service that excludes future uploads from model development after an opt-out. That description would not establish what happened to earlier submissions. Nor would it explain whether the provider retains files for another stated purpose. Do not expand a narrow setting into a promise it does not make.
Turning off a public gallery is also not necessarily a training opt-out. Public visibility and internal processing are separate questions.
Permission and legal obligations need their own analysis
A contractual clause is not, by itself, proof that every use is lawful. Where personal data is involved, applicable data-protection rules matter. The ICO's discussion of individual rights in generative AI highlights issues across training data and model outputs. It is UK regulatory context, not a finding about a particular service or a universal legal rule.
Avoid assuming that consent is the only possible legal basis, or that accepting a website's terms removes all personal-data rights. If the use is consequential or disputed, consult the relevant regulator's current guidance or a qualified adviser.
Deletion is a separate question
Removing a file from an account and reversing any influence on a trained model are not the same operation. A general deletion confirmation should not be interpreted as evidence that a model was retrained or that every relevant system was examined.
Ask the provider to describe the scope of its response. Our guide to upload deletion versus account deletion explains how to keep requests specific without assuming that one button resolves every issue.
A practical question to send support
You can ask whether files submitted through your particular product and account tier are used to train or fine-tune models, which controls apply, and whether those controls affect earlier submissions. Do not attach another private image to illustrate the question.
If the reply simply repeats a slogan, use the privacy policy reading guide to identify what remains unanswered. Uncertainty is a reason to limit what you submit, not evidence that a specific undisclosed practice definitely occurs.
Related reading
What Happens to Photos Uploaded to an AI Service?
