Should we worry if teams are already using ChatGPT without saying so?

Yes, insofar as a Free or Plus account paid for personally gives management no visibility or control over what goes through it. The first step isn't a panicked ban, but finding out who uses it and for what, before deciding on a managed offer or a usage framework.

Do my ChatGPT conversations train OpenAI's models?

We couldn't verify OpenAI's answer on this specific point for ChatGPT: the pages that should answer it (openai.com, chatgpt.com, help.openai.com) refused automated access during this check. What is confirmed is that the "no training by default" policy OpenAI publishes on its developer platform concerns the API, not explicitly ChatGPT.

What's the difference between a personal Plus account and a Team or Enterprise offer?

A Plus account is an individual subscription, paid for and managed by the person who opens it, with no link to the organization that employs them. A Team or Enterprise offer adds, in principle, central administration. We couldn't confirm the details of this distinction on a page accessible for verification.

How much does ChatGPT cost for a team?

We found no OpenAI pricing page accessible to automated verification at the time of writing: neither the price of the Plus offer nor that of a team offer could be confirmed.

What's the difference with Claude or Mistral Le Chat?

The clearest verifiable difference lies in the public documentation: Claude and Mistral publish, on accessible pages, a dated answer to the question of training on consumer-account data. For ChatGPT, this specific answer couldn't be verified under the same conditions.

ChatGPT: our review

ChatGPT is OpenAI's conversational assistant, most often adopted inside a company by an employee who opens a free or Plus account on their own initiative, before management or IT hears about it.

Price not verified

Go to ChatGPT's website

What we take away from it

What counts in its favor

OpenAI publishes a broad set of certifications — SOC 2 Type 2, ISO 27001, 27017, 27018, 27701, ISO 42001 for AI management, PCI DSS — a useful starting point for a security review, once the tool is formalized.
On the API side (so outside ChatGPT itself), OpenAI offers data-localization options by region — European Union, United Kingdom, Canada, Japan, among others — for organizations building on it rather than using the consumer app.

What counts against it

A personal Free or Plus account, paid with an employee's own card, has no admin console: management can neither see what goes through it, nor cut it off remotely, nor enforce a retention policy. That's the very definition of an unmanaged account.
The pages where OpenAI explains its privacy policy for ChatGPT (openai.com, chatgpt.com, help.openai.com) all refused automated access during this check (HTTP 403 codes), even for a simple read. What is confirmed, however, is that the "no training on your data" policy OpenAI documents on its developer platform covers only the API, explicitly not ChatGPT.
No price for the consumer offer could be confirmed on a page accessible for verification: neither the Plus subscription amount, nor that of a Team or Enterprise offer.

The point isn't judging the tool on its technical merits, but dealing with the fact that it's probably already used in your company with no framework: the priority is knowing that, then deciding whether to switch to a managed offer or ban it.

The facts

Data hosting: Hosting not verified
Categories: AI assistants,Writing
Data verified on: 30 July 2026

What happens to your data

Is your data used to train their models?: Not verified
Subprocessors: Not verifiable for ChatGPT itself: OpenAI's consumer-facing privacy pages blocked automated access during this check. What OpenAI documents elsewhere, on its developer platform, concerns the API and explicitly states it doesn't apply to ChatGPT.

ChatGPT in detail

ChatGPT is OpenAI's conversational assistant. In most SMEs we come across, it wasn't chosen: it was adopted. An employee opens a free account or pays for the Plus subscription with a personal card, uses it to draft an email or summarise a contract, and no one else in the organisation knows about it until an incident reveals it.

What it actually does

Everyday use comes down to a few actions: asking a question, pasting text to rephrase or summarise, generating a first draft of a document. The free account and the Plus account are personal subscriptions, opened and paid for by an individual, with no administrative link to the employer. That point, more than the technology itself, shapes everything else in this profile.

Why it interests us — and why it worries us

The question an owner should ask isn't "should we use ChatGPT," it's "who's already using it, and with what data." A personal account has no admin console: management can neither see it, nor cut it off remotely, nor set a retention period. A client document pasted into a chat window escapes the company's control entirely, whatever OpenAI does with it besides. That's the angle we address first in our audit and advisory engagements: mapping actual usage before discussing what tool should replace it.

Where it falls short

Three reservations, and one difficulty that cuts across all of them: verifying what OpenAI states about ChatGPT from the outside is, as of now, hard. The pages where the vendor explains its consumer privacy policy (openai.com, chatgpt.com, help.openai.com) all blocked automated access during our check, even for a simple read. We therefore publish no claim about training its models on Free or Plus conversations: we haven't verified it ourselves.

What we were able to verify, in the developer-facing documentation, is that OpenAI states there that its "no data used for training" policy applies to the API, and doesn't apply to ChatGPT. In other words, even the protection most often cited about OpenAI doesn't cover the product your teams probably already have in their hands.

Finally, pricing couldn't be confirmed on an accessible page: neither the Plus subscription's price nor that of any team offer.

Our recommendation

Don't assume that banning ChatGPT solves the problem: the tool is already in circulation, often outside any software inventory. Start by finding out who's using it and why, train rather than ban in haste, and if the use is legitimate, move it to a company-managed offer rather than leaving personal accounts to handle client documents. Our AI training offer starts precisely from that observation.

Who it's for, and who it isn't

Suited for

Management that wants to first map out real usage in its teams before choosing between banning it, tolerating it under a framework, or switching to a managed professional offer
An organization that has already replaced personal accounts with a Team or Enterprise offer under central administration

Not a fit

Any organization handling data covered by professional secrecy (health, law, finance) as long as usage runs through undeclared personal accounts
A business owner who wants a listed price and a privacy policy verifiable in two clicks: for ChatGPT, neither currently is, from the outside

Frequently asked questions

Should we worry if teams are already using ChatGPT without saying so?: Yes, insofar as a Free or Plus account paid for personally gives management no visibility or control over what goes through it. The first step isn't a panicked ban, but finding out who uses it and for what, before deciding on a managed offer or a usage framework.
Do my ChatGPT conversations train OpenAI's models?: We couldn't verify OpenAI's answer on this specific point for ChatGPT: the pages that should answer it (openai.com, chatgpt.com, help.openai.com) refused automated access during this check. What is confirmed is that the "no training by default" policy OpenAI publishes on its developer platform concerns the API, not explicitly ChatGPT.
What's the difference between a personal Plus account and a Team or Enterprise offer?: A Plus account is an individual subscription, paid for and managed by the person who opens it, with no link to the organization that employs them. A Team or Enterprise offer adds, in principle, central administration. We couldn't confirm the details of this distinction on a page accessible for verification.
How much does ChatGPT cost for a team?: We found no OpenAI pricing page accessible to automated verification at the time of writing: neither the price of the Plus offer nor that of a team offer could be confirmed.
What's the difference with Claude or Mistral Le Chat?: The clearest verifiable difference lies in the public documentation: Claude and Mistral publish, on accessible pages, a dated answer to the question of training on consumer-account data. For ChatGPT, this specific answer couldn't be verified under the same conditions.