AI training and your files

    Who trains AI on your files? What twelve services say in writing

    Before you upload a contract or a client file, it is fair to ask what the service is allowed to do with it. We read what twelve file sharing and cloud storage services publish on the question and restated it here, one line each, with a link to their page.

    We make one of these services, which is why nothing on this page is our opinion of the others: it is what they wrote.

    Read on November 12, 2026

    12 services
    Provider pages only
    Read November 12, 2026
    Corrections welcome

    The short answer

    Do file sharing services train AI on your files?

    Whether a file sharing service trains AI on your files depends on the service, and on which of its pages you read. On November 12, 2026 we read what twelve file sharing and cloud storage services publish about it. WeTransfer, Smash, Proton Drive, Tresorit, and OneDrive for work or school state that customer files are not used to train AI. Google Drive, Google Workspace, and Box say not without the customer's permission. pCloud and ShareFile say it for one feature only.

    Dropbox's privacy policy FAQ describes models for search, organisation and summaries that may be trained on your documents and metadata, while its AI principles say it will not build generative AI models from customer content without consent. Microsoft's privacy statement says it may use your data to train its AI models, in a sentence that does not name OneDrive. For Filemail and Sync.com we found no statement about AI training. Every line in the table below links to the provider's own page.

    In their own words

    What each service says in writing

    Each line restates a page published by the provider and links to it. Read on .

    They state that files are not used to train AI
    WeTransferIts content protection principles say it does not use your content to train AI models.Source: Principles of Content Protection
    SmashA help article on GDPR says no AI training is done with your files.Source: Is Smash GDPR compliant?
    Proton DriveThe Drive page promises no ads, no AI training and no scanning.Source: Proton Drive
    TresoritThe SecureCloud page says Tresorit does not analyse customer files or train models on them.Source: Tresorit SecureCloud
    OneDrive, work or schoolMicrosoft's documentation says none of an organisation's data is used to train AI models, large language models or any other models.Source: Privacy, security, and compliance in Microsoft OneDrive
    Not without your permission
    Google DriveA Drive help page says Google does not use Workspace data to train or improve the generative AI and large language models behind Gemini, Search and other systems outside Workspace without permission.Source: How Gemini in Workspace protects your data
    Google WorkspaceThe Workspace privacy hub says customer data is not used to train models without the customer's prior permission or instruction.Source: Generative AI in Google Workspace Privacy Hub
    BoxIts AI FAQ says Box does not train on customer data, that its vendors do not train on data sent from Box, and that Box AI will only train on data with customer approval.Source: Box AI Frequently Asked Questions
    They say it for one feature
    pCloudA help article says the optional Photos AI feature does not use your photos, or what it generates, to train AI models. We found no statement covering all stored files.Source: pCloud Photos AI
    ShareFileIts AI FAQ says the secure share recommender does not use file content to train the underlying AI. Its End User Agreement also addresses AI products.Source: ShareFile AI capabilities FAQ
    Their policy describes training on user data
    DropboxIts privacy policy FAQ says Dropbox may build models that identify keywords and topics in a document, that these models may be trained on your documents and metadata, and that they power search relevance, automatic organisation and document summaries. Its AI principles say it will not build generative AI models from customer content without consent.Source: Dropbox Privacy Policy FAQ, Dropbox AI principles
    OneDrive, personalMicrosoft's privacy statement says it may use your data to develop and train its AI models. That sentence does not name OneDrive, and we did not find a Microsoft page that settles the question for files in a personal OneDrive.Source: Microsoft Privacy Statement
    No statement found
    FilemailIts privacy policy does not mention AI, machine learning or model training.Source: Filemail Privacy Policy
    Sync.comIts privacy policy says it does not access or process the files you store, apart from listed exceptions. We found no statement about AI training.Source: Sync Privacy Policy
    And Sandra
    SandraOur privacy policy says we do not use your files, transfers or Vault documents to train AI models, ours or anyone else's, that no AI system reads them, and that we do not hand them to anyone for that purpose. Our terms limit our licence on your content to storing and transmitting it.Source: Privacy Policy, Terms of Service, Security

    Terms change. Open the source before you rely on a line, and tell us through the contact form if one is out of date: we will correct it.

    Method

    Only what the provider itself publishes

    For each service we opened its own pages: terms, privacy policy, help center, product and trust pages. We looked for one thing, a statement about using the content of customer files to train AI or machine learning models.

    We did not use press articles, forums or summaries, and we did not deduce anything from general licence wording. When a provider says nothing explicit, the line says so.

    A missing statement is not an admission. It means the question is not answered in writing.

    Reading the table

    Three things the lines do not tell you

    Account type
    A provider can commit more firmly for business accounts than for personal ones: the two OneDrive lines above are an example. Check the line that matches the account you actually use.
    Where it is written
    A sentence in a contract or a privacy policy binds the provider more than the same sentence on a product page. The source link shows which one you are reading.
    Features you switch on
    A commitment about stored files may not cover an AI assistant you choose to connect, or data you choose to share with one. Those have their own terms.

    In practice

    Keeping a file out of AI training

    01

    01

    Look for the sentence

    Before a sensitive upload, find the provider's statement on AI training. If you cannot find one, ask their support and keep the answer.

    02

    02

    Check the settings

    Turn off AI features and data sharing options you do not use, on personal accounts especially.

    03

    03

    Encrypt before upload

    With end-to-end encryption the provider stores a file it cannot read, whatever its terms say.

    Common questions

    Common questions

    Does WeTransfer use my files to train AI?

    WeTransfer's content protection principles say it does not use your content to train AI models (page read on November 12, 2026).

    Does Dropbox use my files to train AI?

    Dropbox's privacy policy FAQ says it may build models that identify keywords and topics in a document, and that these models may be trained on your documents and metadata to power search, organisation and summaries. Its AI principles say it will not build generative AI models from customer content without consent (pages read on November 12, 2026).

    Does Google Drive use my files to train AI?

    A Google Drive help page says Google does not use Workspace data to train or improve the generative AI and large language models behind Gemini, Search and other systems outside Workspace without permission (page read on November 12, 2026).

    Does OneDrive use my files to train AI?

    It depends on the account. For work or school accounts, Microsoft's documentation says none of an organisation's data is used to train AI models. For personal accounts, Microsoft's privacy statement says it may use your data to develop and train its AI models; that sentence does not name OneDrive, and we did not find a Microsoft page that settles the question for files in a personal OneDrive (pages read on November 12, 2026).

    What does it mean when a service says nothing about AI training?

    Only that the question is not answered in writing. It does not show that the service trains AI on files. Ask its support and keep the reply.

    Does Sandra train AI on files?

    No. Sandra never uses your files to train AI models, and no AI system reads them. With end-to-end encryption on, Sandra stores only ciphertext.

    How was this page made, and how do I report an error?

    Each line restates a page published by the provider itself, opened on November 12, 2026 and linked. If a line is out of date, write to us through the contact form and we will correct it.

    Send a file that stays yours

    No account needed. No ads, no AI training.

    Send a file