privacy

Photobucket Is Selling Your Old Photos to Train AI

Photobucket quietly updated its privacy policy to allow selling user photos for AI training. A class action followed. Here's what it means for your old photo accounts.

Photobucket was one of the original photo hosting platforms of the internet. At its peak, it stored billions of images for hundreds of millions of users. Many of those users moved on to Google Photos, iCloud, Instagram, or other platforms — but they did not delete their Photobucket accounts. The photos are still there.

In late 2024, Photobucket updated its terms of service to allow the company to share user images with third-party companies for AI training purposes. A class action lawsuit followed in late 2024 and is ongoing. The case has drawn renewed attention in 2026 as the AI training landscape has intensified and legal scrutiny of such policies has grown.

For anyone who created a Photobucket account in the 2000s or 2010s and has not thought about it since, this is worth your attention.

What Photobucket Changed in Its Terms

The terms update, filed with the court in the class action complaint, allows Photobucket to sell or license user photos to third parties — including AI companies — for the purpose of training machine learning models.

The change was made without individual notification to users. Photobucket updated the terms on its website and notified the change through a general update notice, consistent with what its terms allow it to do. Users were not individually emailed. They were not asked to opt in.

The practical effect: photos you uploaded to Photobucket at any point in its history — including decades-old family photos, personal portraits, candid shots from your past — may be used to train AI image models if you have not deleted your account and opted out.

Why Abandoned Accounts Are the Core Problem

Photobucket’s situation is the clearest example of what privacy researchers call the “ghost account” risk. Millions of people have accounts on platforms they no longer use, with data they no longer think about, governed by terms they have not read in years.

When those platforms update their privacy policies — as they periodically do, often to expand their rights to use your data — the update applies to all accounts, including dormant ones. The user who stopped uploading to Photobucket in 2012 has the same new terms as the user who logs in every week.

Ghost accounts are not rare. Most people who used the internet in the early 2000s have accounts on platforms that no longer serve a function in their digital lives. Each of those accounts holds data under terms that can be changed.

The Class Action and What It Argues

The class action lawsuit filed against Photobucket makes several claims. The core argument is that using photos for AI training purposes without explicit user consent exceeds what users agreed to when they created their accounts under the original terms, and that a unilateral terms update after the fact does not constitute adequate consent.

The lawsuit also raises questions about the financial benefit Photobucket receives from licensing photos versus the lack of compensation — or even notice — to the people whose images are used.

As of this writing, the case is in discovery. No liability finding has been made. Photobucket has not publicly conceded any wrongdoing.

Legal outcomes in cases like this are uncertain and slow. The more immediate practical question is what you should do now, regardless of how the lawsuit resolves.

The Broader Pattern: Old Platforms Monetising Your Archive

Photobucket is not unique. Several platforms have moved to expand data monetisation rights in recent years, often framing the change around AI development:

  • Adobe updated its terms in 2023 to assert broader rights to user content for training AI features in its products. After significant backlash, it issued a clarification that it would not train AI on user content without a separate agreement — but the original terms change caused a large user response.

  • Flickr and similar photo communities have faced similar pressure as the AI licensing market has developed.

  • Reddit and Twitter/X have licensed their full content archives to AI companies, affecting any images users posted through those platforms.

The connecting thread is that platforms accumulate large archives of user content, that content has become commercially valuable as AI training data, and the terms allowing its use are often updated well after the original relationship was formed.

What to Do About Your Old Photo Accounts

Audit what you have. Most people have accounts on more platforms than they can immediately recall. Try to remember every photo hosting or social media service you have used — Photobucket, Flickr, Picasa (now shut down), Snapfish, Shutterfly, old forum accounts, early social networks. Each is a potential repository of images you have not reviewed in years.

Delete what you do not use. If you no longer want an account to exist, delete it — not just deactivate it. Most platforms keep deactivated account data; only deletion triggers the data removal process, often subject to a grace period.

Download before you delete. Before removing any account, use the platform’s data export tool to retrieve a copy of everything you stored there. This is particularly important for accounts with photos you would want to keep even if you have no interest in continuing the relationship with the platform.

Read terms updates when they arrive. Most platforms notify users of terms changes by email. These emails are easy to ignore — they are not written to be read. But a terms change is the moment to decide whether you still want to be a customer.

Check the opt-out options. Photobucket, in response to the lawsuit, now offers an opt-out from the AI training data provisions in its updated terms. The mechanism for exercising this opt-out is in the account privacy settings. Other platforms with similar provisions have added similar opt-outs under pressure. The option is not always prominently advertised.

The Specific Concern With Old Personal Photos

Photos uploaded to early hosting platforms were often more personal than what people share today. Early Photobucket users were not curating a public social media presence — they were storing photos for friends and family, or hosting images for personal blogs and forums. The content was often candid, unguarded, personal.

Those images can now be part of AI training datasets. The concerns this raises are not purely abstract:

Facial recognition training. AI models trained on large photo datasets learn to identify and reproduce faces. Personal photos from your past may contribute to models capable of generating convincing likenesses of private individuals.

Personal moment reconstruction. AI image generation has advanced to the point where training data can influence the specificity of generated outputs. Your personal moments become reference material for a technology whose outputs you have no control over.

No timeline on use. Once images are incorporated into a training dataset, their contribution to the model is not easily reversed even if you later delete the original account.

Where Your Older Photos Are Safer

The lesson from Photobucket is not that any cloud storage is inherently unsafe. It is that the terms under which you store personal data matter, and that platforms whose business model depends on monetising that data — through advertising, licensing, or AI partnerships — have ongoing incentives to expand what they are permitted to do with it.

A privacy-focused personal storage service operates under a different business model. daftei’s stated policy is that it will never sell user data to any third party, never train AI models on user content, and never show ads. Storage is free up to 5 GB, with Pro plans starting at $5.99 per month or $44.99 per year. There is a lifetime plan at $89.99. Files are stored with AES-256 encryption at rest and TLS 1.3 in transit, on a service that is GDPR and CCPA compliant.

The difference is not just technical. It is structural: when the business model is storage, the business does not benefit from using your data for anything beyond giving it back to you.

One Action to Take Today

Log in to Photobucket — or attempt to — and check whether your account exists. If it does, use the Export function to download anything you want to keep, then delete the account.

Repeat that for every photo hosting platform you can remember using. It takes a few hours and eliminates the ghost account risk entirely for those platforms.

You cannot undo training that has already happened. But you can prevent future data from being added to whatever licensing pipeline these platforms have built.

The class action may or may not succeed. The habit of auditing your old accounts and closing the ones you no longer use protects you regardless of the legal outcome.

Your memories deserve better than an ad platform.

Try daftei free →
← All posts