You finish a project, zip up the files, and send them through a file-transfer link because the attachment is too big for email. It feels like the most mundane action in your digital life — a courier service for bytes. In 2026, it stopped being that simple.
Earlier this year, WeTransfer quietly updated its terms of service to include language granting itself a “perpetual, worldwide, non-exclusive, royalty-free, transferable, sub-licensable license” to use uploaded content — including, critics pointed out, to improve machine learning models. The backlash was fast and loud enough that WeTransfer reversed course within days, clarifying that it does not and will not use customer content to train AI. But the episode revealed something most people hadn’t considered: the file-sharing link you use to send a contract, a family video, or a portfolio might come with terms that say more than you think.
Why File-Sharing Services Are Different From Storage
Cloud storage and file-transfer services often get lumped together, but they’re built around different default assumptions.
A storage service’s pitch is “keep this here long-term.” Its terms of service usually address retention, access controls, and what happens to your account over time — because the relationship is ongoing.
A transfer service’s pitch is “move this from A to B, then it’s done.” Historically, that framing made people less likely to scrutinize the terms — it’s a courier, not an archive. But the files still pass through the provider’s servers, get processed, and in many cases sit there for days or weeks before the recipient downloads them (or doesn’t). During that window, the same questions apply: who can access this file, what can the company do with it, and does “temporary” actually mean temporary?
The WeTransfer controversy mattered because it made explicit what was previously buried in boilerplate: a “we can use your content to operate and improve the service” clause can be read narrowly (bug fixes, performance) or broadly (training models). Most users never read closely enough to know which interpretation they’d agreed to.
What “AI Training” Clauses Actually Look Like
Few companies write “we will train AI on your files” in plain language. Instead, the relevant rights usually show up as part of a broader license grant. Phrases worth watching for include:
- “Improve our products and services” — broad enough to cover almost anything, including model training, without saying so directly.
- “Sub-licensable” or “transferable” — meaning the company can pass usage rights to a third party, including AI vendors or partners.
- “Perpetual” or “irrevocable” — meaning the right doesn’t expire when you delete the file or close your account.
- “Aggregate” or “de-identified” data use — sounds privacy-protective, but de-identification of file content (versus structured data) is technically much harder to guarantee.
None of these phrases are automatically sinister — most software needs some license to process your files in order to function (compressing them, generating previews, scanning for malware). The distinction that matters is whether the license is scoped to operating the service you asked for or expanded to cover improving systems that benefit the company more broadly, including AI models that other customers or the wider market will use.
The Pattern Isn’t Limited to One Company
WeTransfer was the most visible case in 2026, but it wasn’t isolated. Reporting on the episode noted that Adobe, Zoom, Dropbox, Slack, and other platforms have all revised or clarified AI-related terms after user pushback in recent years — a pattern of “expand quietly, retreat loudly if caught.”
The throughline is that the economics of AI development create a constant incentive to find new sources of training data, and a company’s existing user base — with its enormous volume of real-world files, images, and documents — is an obvious target. Terms of service are the legal mechanism through which that access gets negotiated, almost always in the company’s favor by default, because almost nobody reads them.
This doesn’t mean every file-sharing or storage provider is doing this. WeTransfer’s current position states plainly: “We don’t use your content to train AI models, we never have, and we have no intention of changing that.” The point isn’t that all providers are bad actors — it’s that the terms are where the actual commitment lives, and commitments expressed in marketing copy don’t always match what the legal document says, until someone reads it closely enough to cause a stir.
A Practical Checklist Before You Send Sensitive Files
You don’t need a law degree to do a basic check before sending something sensitive — a signed contract, a video with other people in it, financial documents, creative work you haven’t published yet.
Search the terms for “AI,” “machine learning,” and “train.” Most terms-of-service documents are searchable text. A simple Ctrl+F for these terms will surface the relevant clauses if they exist, without requiring you to read the entire document.
Check how long files are retained after transfer. “Temporary” transfer services vary widely — some delete files after a single download, others retain them for days or weeks by default. If a file sits on a server longer than you expect, it’s exposed for longer than you expect, regardless of what the terms say about training.
Look for an explicit no-training commitment, not just silence. The absence of an AI-training clause is better than its presence, but an explicit statement — like WeTransfer’s revised principles page — is more durable, because it’s something the company can be held to publicly if it changes again.
Consider whether the file needs to leave a “transfer” context at all. If you’re sending something to yourself — a backup of a document, a copy of a video for safekeeping — a transfer link isn’t really the right tool. It’s built for one-time movement between people, not for keeping a copy you control long-term.
Where This Connects to Personal Storage
The file-sharing controversy is really a subset of a bigger question: every time your files pass through a third-party service — to send, to back up, to share — you’re implicitly agreeing to that service’s terms for however long the files sit there. The more services your files pass through, the more terms-of-service surface area you’ve accumulated, often without tracking any of it.
One practical mitigation is reducing how many different services your important files touch. If your personal documents, photos, and recordings live in one place you trust and control, you have fewer terms-of-service documents to audit and fewer copies scattered across transfer links that may or may not have expired.
daftei is built around that principle for personal storage: files are encrypted in transit with TLS 1.3 and at rest with AES-256, daftei does not use your content to train its own AI or any third party’s, doesn’t sell data, and runs no ads. The commitment isn’t buried in a clause that can be quietly expanded later — it’s the basis of the product. With 5 GB free and unlimited storage on Pro, the goal is that your personal files have a home that doesn’t require you to re-read a terms-of-service document every time a company updates its AI strategy.
The Broader Lesson
The WeTransfer episode resolved in users’ favor — public pressure worked, and the company reversed the change within days. But the speed of the reversal is itself instructive: the original change did get published, and would have taken effect for everyone who didn’t notice, if enough people hadn’t noticed.
Terms of service updates happen constantly, almost always without meaningful notice beyond a buried email or a banner that’s easy to dismiss. The files you send and store are not going to stop being valuable training material for AI companies — if anything, demand for real-world data will keep growing. The only durable defense is choosing services where the no-training commitment is explicit, current, and consistent with how the company actually makes money — and minimizing how many services your most sensitive files pass through in the first place.
Keep your files in one place that doesn’t change the rules on you