What Happens to Your Data When an AI Company Shuts Down
AI vendors get acquired, run out of money, or quietly go dark more often than founders like to admit. Here is what actually happens to your data in each scenario, and the checklist to run before you hand a vendor anything you cannot afford to lose.
Your AI vendor emails you on a Tuesday. Subject line: “Important update about your account.” They’re shutting down in three weeks. Every conversation, document, and customer record you fed that tool is sitting on their servers right now, and you have no idea what happens to it next.
Here is what happens to your data when an AI company shuts down: it gets deleted on a fixed deadline, folded into whatever company acquired the assets, sold as part of a bankruptcy estate, or left on servers nobody is paying to maintain anymore. Which outcome you get depends on how the shutdown happens, an acquisition, an asset sale, or a quiet wind-down, and on what you agreed to in the terms of service before you uploaded a file. Most founders only read that document after the shutdown email arrives. By then there's nothing left to negotiate.
How AI companies actually disappear
Founders rarely announce “we are shutting down and deleting everything.” The language is softer: sunsetting, winding down, moving on to new endeavors. Underneath the euphemisms, a shutdown resolves in one of a handful of predictable ways, one more item on the list of ai risks worth understanding before you hand a vendor your data. Here's the pattern, based on how these have actually played out:
Acquihire or licensing deal: a bigger player absorbs the company or its technology, the product often keeps running for a while, and your data typically stays put, at least short term.
Asset sale in bankruptcy: customer data becomes a line item on the estate's asset list. Courts have blocked sales that contradict the company's own privacy policy, but the data doesn't automatically disappear either.
Quiet wind-down: the company sets a shutdown date, usually ten days to a few months out, and expects you to export what you need before the servers go dark.
Sudden dead stop: no notice at all. More common with under-funded startups that run out of runway before they run out of users.
Silent decay: nobody announces anything. Support stops responding, updates stop shipping, and the product quietly rots while your data sits unmaintained.
What this has actually looked like
Humane is a useful case study because it moved fast. In February 2025, Humane sold most of the company, including its software platform and patents, to HP for $116 million and discontinued the AI Pin entirely. Device owners got about ten days' notice: servers would stop responding on February 28, and after that all customer data, including photos, videos, and notes, would be permanently deleted. Owners were told to download what they wanted to keep before the cutoff; refunds only went to buyers inside a 90-day return window. Most walked away with a dead device and nothing else. (PetaPixel)
Compare that with Dot, the AI companion app from New Computer, which shut down later in 2025 with a longer runway. The company announced in early September that the app would keep running until October 5, and pointed users to a “Request your data” option in account settings to export their history before the deadline. Same ending, data deletion and a dead product, but a five-week window instead of ten days. (TechCrunch)
When a shutdown goes through bankruptcy court instead of a quiet wind-down, data becomes a negotiable asset instead of something the company just deletes. Not unique to AI companies: the FTC intervened in RadioShack's 2015 bankruptcy, arguing alongside dozens of state attorneys general that selling customer data as a standalone asset would violate the privacy promises RadioShack made when it first collected that data. (FTC) Courts can appoint a Consumer Privacy Ombudsman to check a proposed sale against the company's own privacy policy, the document most people skim past at signup. That review exists because RadioShack was public and in Chapter 11. A small AI startup winding down quietly, outside of bankruptcy, usually has no ombudsman and no one watching.
Before you sign up: a checklist
None of this is predictable once you're already a customer. The leverage is at signup, before you've put months of data into a tool you can't easily replace. Run through this before you commit, not after a shutdown email lands.
Read the data export section of the terms of service, not just the privacy policy. Confirm you can get your data out in a usable format, such as CSV or JSON, at any time, not only during a shutdown window.
Find the retention and deletion clause. It should cover what happens if you cancel, and separately, what happens if the company itself shuts down. Most terms only cover the first case.
Check for a stated notice period. Some vendors commit to 30 or 60 days' notice; most commit to nothing. Humane gave owners about ten days. Treat silence on this as a risk, not an oversight.
Look for an acquisition or assignment clause. It usually says your data can transfer to a successor company without asking you again. That's normal, but know it's there.
Confirm ownership of outputs, not just inputs. If the tool generates content, code, or models for you, get it in writing that you own the output regardless of the vendor's fate.
Check for a self-hosting or open-weight fallback. A thin wrapper around a closed model with no export path leaves you with nothing if the vendor disappears. A tool you can point at something you run locally gives you a real exit.
Keep your own backups on a schedule, not as an afterthought. Waiting for a shutdown notice to start exporting is how people lose data during a ten-day countdown.
If the shutdown notice already landed
Past the checklist stage and already holding a shutdown email? Move fast. Export everything the tool allows before the deadline, not after. Save a copy of the terms of service page as it exists now, since it may not stay up. Check whether it's an acquisition (data might move to a new owner, worth confirming under what terms) or a full wind-down (it's probably deleted on the stated date, no negotiating). If the tool was central to your workflow, this is also the moment to vet whatever vendor you migrate to against the checklist above. Betting a core workflow on one provider is vendor lock-in you renew with every integration you build on top of it.
FAQ
Do AI companies have to tell you before they shut down?
Not by default. Some terms of service include a stated notice period, often 30 to 90 days, but many say nothing. Humane gave AI Pin owners about ten days between the shutdown announcement and permanent data deletion. Check the termination clause before you sign up, not after.
Can an AI company sell my data if it shuts down?
Sometimes, depending on how the shutdown happens. In a straightforward wind-down, most companies delete data rather than sell it, since selling raises legal exposure. In a bankruptcy asset sale, data can be sold as part of the estate, but regulators including the FTC have blocked sales that contradict a company's own privacy policy. That policy is the document that actually governs this.
What's the safest way to avoid losing data if a vendor disappears?
Export on a regular schedule regardless of whether a shutdown looks likely. Favor vendors that let you run on open-weight models or self-host, which gives you a working fallback instead of a dead account. Avoid feeding irreplaceable data into a tool with no stated export path.
Does data portability law require AI companies to let me export my data?
It depends on jurisdiction and the kind of data involved. Regulations like the GDPR give individuals a right to receive their personal data in a portable format, and similar state laws exist in the US, but these rights generally cover personal data, not business records, model outputs, or proprietary content generated inside the tool. Don't assume a legal right fills the gap a vendor's own export feature should have covered.
How did this land?
About the author

Senior Editor, AI & Product
Cecilia leads the Swarmz editorial desk. She has spent a decade turning complex AI and product topics into writing people actually finish, and she owns the blog's quality bar.


