HELP! OpenAI's ChatGPT Team Plan Just Took All Of My Data As Hostage!
Well, this sure is something else on this corporate rip-off maximizing cyberpunk timeline we suddenly find ourselves on. Also: how to make the horse cry for $200.
TL;DR:
OpenAI’s Team Plan is holding my data hostage. There’s no export option, no merging back to personal workspaces, and customer support suggests manually copy-pasting gigs of data—all while violating GDPR and locking paying users into their walled garden. If I don’t keep paying, I lose a year’s worth of work. OpenAI: stop acting like ransomware, and give me my damn data export.
UPDATE (25 January 2025): I received an email from OpenAI’s staff informing me that they’re looking into this and will be returning back shortly, which implies that this is an actual issue that needs fixing. Meanwhile, this article is still valid.
After around five years of using OpenAI’s products and services as a professional AI model developer. I built my first Finnish GPT-2 model at around 2021 and have been OpenAI’s API customer for nearly five years, building popular AI services like ChatKeke that have utilized their API.
Recently, things escalated to a completely unforeseen point when it comes to OpenAI’s choices in the way they handle and allow data exports and workspace merges between different workspaces (i.e. personal vs. Team).
What happened is almost all too reminiscent of Philip K. Dick’s works to even be believable — basically, what I’m up against at the moment, is my ChatGPT Team Plan suddenly holding my data hostage unless I pay up the annual Team plan fee.
Otherwise, I’ll lose access to all of that data altogether.
This includes all of my chat sessions, projects, you name it, that are in the Team plan workspace.
The most ridiculous thing in this is that this lack of data exports and merges to other plans only applies to OpenAI’s Team Plan, which has been, in lack of better terms, seemingly constructed in gross negligence all business ethics, GDPR and other binding legal regulations.
How did this happen?
It all started a few days back when I received an e-mail from OpenAI’s billing…
Upon receiving that e-mail, I thought to myself: “Well, I’m sure glad they didn’t grab that out of my pocket — if I had any pockets, that is — Bah, don’t need the ChatGPT Team plan anymore... Hah, can’t even afford it. Might as well just stick to the Pro plan since it provides unlimited use and ditch the Team plan altogether — not much purpose in it anyway, now there’s that brand new o1 pro mode in the Pro plan as well — no need to be stacking all of these AI crap subscriptions… phew! Now, just let me get my stuff out of there and I’m all set, good to go, no problem, whatever…”
Then, it suddenly dawned on me.
I can’t export my data out from the Team plan.
“What the —”
Now, for reference, this is the view on the personal plan side (whether you’re using the free plan, Plus, or a Pro plan):
You can clearly see from the image that you can export your data from there, just like that. No problem.
Meanwhile, on the Team plan side:
As you can see, there is no data export option on the Team plan side. I thought, “this has got to be some sort of a screw-up in their UI, somewhere, this can’t be right…”
Then, I try contacting OpenAI’s customer service.
So, wait a minute… Their go-to answer is to merely tell me to copy-paste everything manually from my Team plan to my Pro plan… what the — what drugs are they on in California these days? My chat history might be, due to the codebases and extensive translations that I’ve worked on, upwards to 1–2 GB of data. I’m not doing that with a copy-paste, I’m not that insane (yet).
Data export is a basic feature in almost any SaaS product. Even free tools like Google Docs allow you to export your entire workspace easily — the fact that ChatGPT’s Team plan (a premium service) doesn’t offer this is simply inexcusable.
And meanwhile, the “regular” personal plan in ChatGPT has a data export option, whether it is the Plus or the Pro plan you’re using. And the Plus plan is cheaper than the Team plan, but the Pro plan ($200/mo) is way more expensive than the Team plan. And you can merge your Plus plan with your Team plan, but not the other way around? What— ?!
Well, if you are baffled by all of this, you’re definitely not alone — I found this forum thread from OpenAI’s own forums, with posts starting from February 2024: https://community.openai.com/t/export-chats-in-chatgpt-team/620984/3 — and as of writing this, in January 2025 — no news.
In that OpenAI community forum thread, there are comments that are near-identical to my rambles here: it’s a GDPR violation not to have data exports — and, how it’s an incredibly bad move from OpenAI — “very problematic”, someone notes. Some are even trying to come up with their own batch download scripts for Team plan workspaces, then, there’s a summer silence — and someone says: “This place is a ghost town.” — “Is there anymore details someone can provide?”
UPDATE (Jan 14, 2025): I’ve opened up a new thread about the issue here: https://community.openai.com/t/urgent-gdpr-violation-no-export-option-for-chatgpt-team-workspace-data/1090657
“What are thou doing, OpenAI?”
It’s odd, to say the least. Even if you don’t care about GDPR or even your own intellectual property, or whatever you might call your own drivels, this is still coming back to a really bad business approach.
Among plausible explanations could be i.e. that they developed and rolled out their different subscription plans and workspaces a bit too hastily — or, maybe they were just a bit too high while at it (on hypium, at the very least) — and thus, ended up shipping all these half-baked trap-plans and workspace “solutions” to customers left and right, causing this whole mess.
That still doesn’t excuse the fact that the absence of data exports on the Team plan side is clearly violating GDPR and other regulations. It’s effectively holding customer data for ransom. You don’t have to be an EU citizen to see exactly what’s going on here. Everyone’s getting the same squeeze no matter where you are and how much you’ve paid for it.
Sadly — and I’d rather not think about this option — perhaps it was intended this way by OpenAI all along — to keep their high-paying users locked in — in which case, that’s just a bleak observation and sets a new low for OpenAI altogether.
It would imply OpenAI profiteering over this on purpose by keeping their paying customers along with their projects captive in a costly plan like that. It’s a troubling thing to even consider: it’d mean that the Team and Enterprise plans are essentially giving their customers a “Hotel California” treatment: paying users have their chats, instances and thus data locked into their Team plans with no export option at all.
Still, at least in the ChatGPT UI, OpenAI still openly encourages their customers to i.e. merge the data from their personal workspace to their Team plan one — and little do many people realize (until it’s too late), that merging their workspace or otherwise engaging with a Team plan workspace seems to be a completely one-way street.
When the ChatGPT Team and Enterprise plans were first introduced, they were advertised on purpose for heavy users of OpenAI’s products — something that would get you higher usage caps, etc — that’s what OpenAI did — they promised higher usage caps for Team plan users. That was the plan and the promise, wasn’t it?
And yet, this is the outcome.
You can only merge your personal plan (free, Plus or Pro) and the contents within into the Team plan, and all of your chats and all of your data will stay there. You cannot get it out via any regular export methods.
“Stop right there, criminal scum! Nobody breaks the law on my watch!”
Not to take this issue all the way to Cyrodiil, but unless I’m totally mistaken, by not having data exports (even most social media platforms have these), OpenAI’s Team plan is currently in gross violation of i.e. EU’s GDPR legislation, which clearly states:
a) Article 15 (Right of Access): Under GDPR, you have the right to obtain a copy of all personal data your organization holds concerning yourself.
b) Article 20 (Right to Data Portability): You are also entitled to receive this data in a structured, commonly used, and machine-readable format (e.g., JSON, CSV).
You can see from the earlier image that showed the “Data controls” in OpenAI’s ChatGPT Team plan, how easy it is to merge data from your personal workspace to the Team plan workspace… and if you do that, that’s it then — you’re screwed. No data exports for you anymore, buddy.
In other words: the merging of workspaces only works one way. And by then, they’ve got you and your data trapped into their locked-in walled garden.
Yes. Once your data is in a Team or an Enterprise plan workspace, you can’t export it out anymore, at least not with OpenAI’s help.
Again: you’ve read it right — as it currently stands, it’s a downright trap.
And you’ll be paying for it, because OpenAI is holding all of your sessions and the data within as hostage — you’ll lose access to whatever is in that workspace if you don’t pay up the separate Team plan bill. Even if you have the $200/month Pro plan!
I have been sending numerous messages to OpenAI’s help center on this issue, to no avail. “Would be a pity if you lost access to all of your data…” — what kind of a team comes up with stuff like this and still thinks it’s somehow done with “an ethical framework” and is “democratizing AI”?
Let’s call a spade a spade: that’s locking professional and heavy user type customers in and then subsequently blackmailing them.
I don’t even know what should I try next. Leave a GDPR violation report on OpenAI’s actions to our data authorities back here in the EU? Try to rally up people across the world (since this affects OpenAI’s Team and Enterprise plans worldwide), to petition legislators and authorities to look into OpenAI’s GDPR policies, just because they refuse to have a data export functionality on the Team plan?
If you look at OpenAI’s privacy policy, they explicitly mention the right to data portability as required by GDPR:
"You may have the right to... Transfer your Personal Data to a third party (right to data portability)."
This is in direct conflict with the lack of a data export mechanism for Team plan users. OpenAI cannot legally deny this right if the data in question constitutes personal data under GDPR. The Team plan essentially holds user data hostage without providing a means for export in a structured, machine-readable format. This violates GDPR Articles 15 (right of access) and 20 (right to data portability).
OpenAI also claims, on the very same privacy policy page:
"We’ll retain your Personal Data for only as long as we need in order to provide our Services to you, or for other legitimate business purposes..."
However, the inability to export Team workspace data more than suggests a one-way retention mechanism. Once the data is in the Team plan, there’s no practical way to retrieve it — especially in larger data dumps, even though OpenAI is obligated to provide access to and portability of that data upon request.
What the hell were they thinking at OpenAI when they came up with this crap? Is it really their view that a premium-paying customer doesn’t have a right to their own data once it has entered OpenAI’s systems? Why is the data export option available in the Plus and Pro plans, but not in the Team plan, when the Team plan is more expensive than the Plus plan? And the Pro plan is way more expensive than the Team and Plus plans combined… and yet, there is no data export in the Team plan. And as to data merging, you can only merge it one way.
None of this makes any sense at all.
Absolutely ridiculous.
And I’m paying premium to get fleeced and screwed over like this?
With my own data?
Fool me once… twice … shut up and take my money —
Jokes aside; this is an absolute cyberpunk/PKD level nightmare.
OpenAI is holding a year’s worth of my own personal work as hostage right now.
OpenAI’s message is, to me and all the other Team plan users worlwide: “Either pay up the annual Team plan bill (that is completely useless for any Pro plan user at this point), or lose access to all of the chat sessions you’ve ever had inside the workspace.”
As to OpenAI’s help center, they either just stonewall me or copy-paste the same go-to answer, no matter how I form my request: “ just copy-paste it out if you want it out of the system.” Yeah, gee, thanks a lot. I’ll do that with a year’s worth of work, could be 1, maybe 2 gigabytes, no biggie, I’ll just start copypasting when it could’ve been done with a single automated click in the personal workspaces under the Plus or even the Pro plan. This is absolutely bonkers.
If the whole thing has been done like this on purpose, it’s even more incriminating: it reeks of a “lock-in” practice where OpenAI wants you to be permanently locked inside their walled garden, and literally hold your data hostage, and the more I look at it, the more it seems that way, that it’s not a bug, it’s a feature. Now, as many of us know, these kind of SaaS walled garden-type designs are often horrible by default, I’d call them predatory even, but even in almost all of those instances, by any sane standards, you still can get out of them with your own data intact.
The ChatGPT Team plan data export policy is blackmail as it stands. You want your data out? You’re left with either setting up a hacky web crawler or an AI agent, completely on your own, which either does the job for you or gets you banned for trying to crack through the “Ubik” door to get your own damn junk. Another option is to just do it manually with copy-paste, as suggested by OpenAI’s help center.
Or, just accept that your data, everything you’ve worked on during that billing year, has been swallowed up by OpenAI’s insatiable Moloch’s Maw, and there’s not a damn thing you can do about it.
And that’s in clear violation of, at the very least, the EU’s GDPR, and possibly many other international legislation as well. It’s not an EU thing — it affects all Team plan users worldwide.
How can you trust these people with anything if they literally come up with systems that swallow up your data unless you pay up?
Also, what’s next? What if what’s seen right now in OpenAI’s Team plan data export blockade sets a new moral precedent on how things are handled by default? We’re living in an era of walled gardens, and now it seems that the walls are constantly closing in on all of us.
Luckily for me, I dodged the bullet of merging workspaces, which would’ve meant an even bigger “data hostage” situation — but a significant portion of my personal works and countless lines of code and text still remain inside my Team plan workspace, and I can’t get it out unless I somehow automate i.e. a separate AI agent to go through all of my chats and get all the data out, chat-by-chat.
This is frankly insane, and I have left multiple written protests to OpenAI’s help center on this, implicating that they’re more than likely violating not just GDPR, but multiple other regional business practices and laws, by not giving whoever is paying for the service a chance to export their data in any reasonable manner.
I’ve sent a written notice to OpenAI that continued denial of a legally compliant export mechanism risks significant repercussions under the GDPR, including potential fines of up to €20 million or 4% of global revenue, whichever is higher.
Although this matter concerns compliance within the EU, it may also raise concerns among customers in other regions who face similar difficulties retrieving their data.
If OpenAI’s current policies effectively deny or severely hinder these rights, they’re on shaky ground legally, especially in the EU. It doesn’t help that Team plan is pitched to heavier or commercial users: that might even elevate the severity from a regulatory standpoint.
Imagine what kind of future this paints. The picture should be crystal-clear: your AI-assisted code editor, your text editor, your collaborating AI can just swallow up all of your work one day — go like the door from Philip K. Dick’s “Ubik” and demand more money or it won’t open up — and if you try to force your way through, you’ll get flagged for a ToS violation. Welcome to the Brave New World, now pay up if you want your daily dose of soma. Whoops, better not be giving Silicon Valley techbros any more ideas.
This isn’t democratizing AI, it’s locking up people, their data and/or their money in an AI dungeon, and I don’t mean the unhinged text-RPG platform from years ago.
That’s not to say that OpenAI couldn’t still fix all of this, here’s how:
OpenAI needs to roll out a fully functional data export tool for Team plan workspaces. This should allow users to download all their data in a structured, machine-readable format (JSON/CSV) with one-click simplicity—just like what’s available on the Personal, Plus, and Pro plans. No half-measures, no "someday"; this is GDPR compliance and basic business ethics.
Allow seamless merging back into Personal/Pro plans or other workspaces. Data flows shouldn't be a one-way trip into OpenAI's walled garden. Users must have the ability to transfer their projects, sessions, and work without arbitrary restrictions.
OpenAI needs to clarify why this isn’t already in place. Was it intentional? A design oversight? They should provide a roadmap and timeline for fixes, making this a public priority.
Those would be my top three suggestions to fix the issue, but again, I’m not very optimistic about anything happening on that frontier — if you look up online discussions, they have locked in users using the Team (and possibly Enterprise) plans like that for almost a year now, ever since the Team plan was rolled out, with no fix in sight.
We’re on a really weird timeline right now, and no, I don’t like it a single bit.
Anyway, moving along with our regular AI updates… *sigh*
“Make The Horse Cry For $200”
(Or €229/month if you’re in the EU. Sora not included.)
Well, now that we got that one blatant rip-off out of the way, if you’re still there, alive and screaming, I might as well recap the other latest developments from OpenAI in here — and I’d hate to get all Gary Marcus when it comes to shaking one’s head in disbelief towards AI development and where it’s heading, so, let’s try to recap the good, bad and the ugly, shall we?
You remember that time when OpenAI’s CEO Sam Altman said something about AI progress heading towards “intelligence too cheap to meter”?
Well, that whole statement got a bit of a rewrite in late 2024.
Deviating to a whole new trajectory from their GPT-4-series (GPT-4, 4o, 4o-mini), OpenAI introduced its new “reasoning models”, the “o”-family, the letter supposedly standing for “Orion” — it had been rumored that OpenAI had this thing called “Project Orion” in the works before the first o1 models were introduced sometime last year — o1-preview and o1-mini, both of which were available for Plus and Team users to tinker around with in what seemed to be a controlled pre-release kind of way.
Out of these new “reasoning models”, o1-preview seemed highly promising, but you could only get somewhere around 50 queries per week with it - whether you had the Plus or the Team plan. Altman promised to lift that usage cap to higher figures as time goes by, but so far there hasn’t been much signs of that.
I enjoyed o1-preview, whereas the smaller o1-mini was just completely off the rails straight out of the box. Sure, it could work on something small, but whenever it got an “idea” on how something was supposed to be, it would just spin it infinitely, like a maniacally fixated person that you couldn’t talk out of their obsessions. I even had to add separate instructions for the model: “YOU STOP WHEN I SAY STOP” — and even this didn’t stop it. Imagine if that thing had been running heavy machinery, or whatever industrial automation… maybe we won’t go there. Too doomy.
Then, with OpenAI’s “Shipmas 2024”, people got a bunch of new AI toys to tinker with, almost like they had their own little advent calendar — counting down the days to Christmas, day by day, there were new roll-outs: Advanced Voice Mode, Sora, the actual o1 release version — an on top of the actual o1, an even better “reasoning model” called “o1 pro mode”.
To access the o1 pro mode, you needed to have the brand new OpenAI Pro plan, which I briefly mentioned before — and with o1 pro mode, it wasn’t just the model being something completely different — its pricing was just that as well.
I immediately named the Pro plan as the “Make the horse cry for $200” plan, or €229 EUR in the EU (almost $250 with current exchange rates as of writing this), with Europeans having no access to Sora, the new image generation model, yet still paying de facto more than US customers, since EU taxes are added on top of the €200 price tag.
Anyway, this whole-new “Make the horse cry for $200” ChatGPT Pro plan was announced in December 2024 — and if you’re wondering, the title is a reference to an old, crude joke that popped up in o1-preview’s chain-of-thought window in someone’s X post when asked to come up with the world’s funniest joke. Now — I will not repeat the joke here for work safety’s sake, that’d just be too funny — let’s just leave it at that.
From Strawberry To Dingleberry, a.k.a. “Guess what the first ‘S’ in ‘SaaS’ stands for?”
OpenAI CEO Sam Altman, in his typical — how should I put it — nonchalant manner, posted on Elon Musk’s X (former Twitter) that he chose the $200 price tag himself. Then he kept on flexing his gold-plated credit card in another X post, admiring its beautiful design.
It’s almost as if the guy’s trying to tell us something.
I wonder what that might be.
Anyway, with the Pro plan’s promised “unlimited usage” (* within “reasonable” use, or however they put a limitation of something that was advertised as “unlimited”), of course I thought I wouldn’t be needing a Team plan for anything anymore.
And now I’m in this mess, where I can’t get my data out from my damn ChatGPT Team plan.
But before we let us become too upset, let’s look at the bright sides: in comparison to the Plus and Team plans, the Pro plan (supposedly) hasn’t got any usage limits considering any of the models that are available, whether it’s GPT-4o, o1-mini, o1, or o1 pro mode. Altman says OpenAI is losing money with the Pro plan. Many in the AI industry say that means “we’re about to go on a price hike” in non-corpo talk English.
Although I hit my own Pro plan rate limits within a day or two of using o1 Pro Mode. By screaming at OpenAI’s help center in my usual unhinged manner, I got my rate limits lifted. “Whoops, our bad!”, they said. “Thanks a lot”, I replied.
Life sure ain’t always easy, and these systems don’t always make things easier.
You know, I sometimes wonder what would have I done with all the money, time and effort I’ve poured into OpenAI’s products and services over the years, as you can see i.e. from my GitHub page (https://github.com/FlyingFathead), many of my repositories utilize either OpenAI’s API’s somehow (i.e. different ChatKeke instances) or OpenAI’s other products, such as their local Whisper models.
But it’s just odd. So many of my works are tied to OpenAI in one way or another. I’ve had their Plus plan, their Team plan, their Pro plan, am a tier 4/5 API user. Still, it’s as if none of it matters, at least to them. The squeeze just never stops. Even with these ideologically driven people in the AI field who claim to have a moral compass that they abide by, money is always the bottom line. If you want tools instead of trinkets, suddenly, the price is so steep that most mortals can’t afford it.
I wonder if the next frontier model will cost $2000/month, given the trajectory we’re on?
Who’ll use it at that point?
And for what?
I’d of course have to keep things objective and not call OpenAI’s ongoing developmental arc as a constant barrage of bait-and-switches and price hikes, since we did get the GPT-4o-mini model after all, and, although it was a bit of a “side-grade” from the good’ole trusty GPT-3.5, even when its price went into a fraction of the old costs, and especially with prompt caching and other cost-cutting measures on OpenAI’s end, the prices have gone down for i.e. API users, but still.
Is OpenAI just poking where’s the ceiling on what people are willing to pay for their services, or what kind of service are they willing to tolerate, or even what and how much they can afford?
In many places, the reality is anything but some sort of Silicon Valley hopium: common folks aren’t always doing all that well financially, even in the so-called first world nations. We’ve got crumbling infrastructure, societal structures altogether are on a shaky foundation, and then these new frontier AI models are compute and power-hungry as hell…
As I’m writing this, there’s a headline in Forbes: Klarna CEO says he feels ‘gloomy’ because AI is developing so quickly it’ll soon be able to do his entire job — imagine a near-future hobo, asking if you got any loose crypto that you could donate to them over the smartphone: “Hey, buddy, can you spare me a bit’o‘pute? You know, back in the day, when they still hired us meatbags, I used to be a CEO… now, look at me…”
What a world.
I know, “I’m getting all existential”, I suppose. It’s just this never-ending SaaS garbage (or is it more and more becoming “AaaS”, as in, “AI as a service”?) … Well, option B is to run your own crap model somewhere and deal with all of that then. I guess it’s more or less a “pick your poison” type of scenario.
Although I’m impressed that o1 pro mode can even code a bit of low-level stuff like assembly, and even interpret from disassembled code, which is cool, I admit — it’s still… it’s like something is off with the bigger picture.
As a somewhat hilarious anecdote amidst the bleakness: one of the o1 models even suggested that if the data export is not available, maybe we should just figure out some sort of a hack for OpenAI’s session cookies to get the chat sessions out from the Team plan. Perhaps we don’t do that, buddy, but thanks for offering “creative solutions” nonetheless.
Maybe there’s just a diminishing return in everything, maybe the joke is getting old. Whether it’s Altman’s oddball flexes of a gold-plated credit card when people are screaming about OpenAI’s pricing and overall AI ethics, or someone like Elon Musk cringeposting his latest modern-day Louis XIV takes on his X platform while sky-high on horse anesthetics at 3AM... hard to tell.
Hmm. Although this article started as a Kafkaesque-P.K.Dickian ramble on how quickly you’ll get rug-pulled in the AI field, it’s not just OpenAI threatening to lock up a year’s worth of work of mine. It’s obviously sad that there’s probably a lot of ideas in that workspace that could still be refined if I rummaged through all of those piles of data, but at this point… it’s just overall melancholy that’s settling in.
How did we end up into this?
I thought OpenAI’s mission was to democratize AI, not to manifest these types of existential hellscapes where you’re stuck with a machine that just swallows up a year’s worth of your work if you don’t pay up.
In terms of technology, the other leading AI companies aren’t doing much better either when it comes to OpenAI. Not much to say about them at this point; with OpenAI’s o1 pro mode still being pretty good, all the bashing here aside, Anthropic’s Claude is still goody-two-shoes, Google’s Gemini remains a post-Bard ridiculous contraption — go ahead and ask Gemini’s 2.0 Flash, “how many r’s are there in ‘dingleberry’?” — you’ll probably get an answer between 3–5.
And so, as you can see, they solved the “strawberry problem”, but the “dingleberry issue” still persists.
-H
The writer is pouring his last pennies into OpenAI’s products and services; all traces of sanity having quit the circus a long time ago.