What changed on 24 August
OpenAI extended advertising in ChatGPT to 31 European countries, including France, from 24 August 2026, following a test launched in the United States earlier in the year. The ads concern the free and entry-level plans; the higher paid plans and enterprise offers remain ad-free. OpenAI says the ads are kept separate from the model’s answers and, in Europe, are not personalised in order to comply with the GDPR.
A signal more than a detail
You could see these banners as just another business model. Above all, they are a reminder: a consumer assistant is first and foremost a consumer product, designed for the widest audience and funded accordingly. Its priorities (audience, engagement, monetisation) are not those of a company entrusting its documents to a tool.
Yet in practice, a large share of professional AI use still goes through these free accounts, opened personally by employees. That is the heart of shadow AI: company data processed in an environment the company does not control, and which now displays advertising.
Different plans, different guarantees
The arrival of ads highlights a little-known reality: behind the same name, consumer AI tools come in several plans with very different guarantees. Terms of use, retention of conversations, possible use to improve the models, presence or absence of ads: everything depends on the plan. In a company, each employee often uses their own, sometimes free, sometimes personal.
The result: the company knows neither which tools are used, nor under which conditions its data is processed. The same question asked by two colleagues may fall under two different contractual regimes.
The questions to ask
- Who in your teams uses an AI assistant, and with which account?
- What data is pasted into it: client documents, contracts, HR data?
- What happens to those conversations: are they kept, and where?
- Can the tool change without you: new terms, new features, new business model?
These questions are the starting point of any serious governance of generative AI, and the first part of an AI audit.
What enterprise AI must guarantee
- You are the client, not the product: no advertising, no exploitation of your conversations.
- Your data stays with you, with no transit through third-party servers.
- A predictable cost, independent of the number of requests and of the plan chosen by each employee.
- A tool connected to your documents, for sourced rather than generic answers.
Offer something better than the free tool
Banning is not enough: employees come back to the tool that saves them time. The only lasting answer is to give them an assistant at least as capable, but designed for the company. That is what MIA does: generative AI installed on your servers, connected to your documentation, ad-free and with unlimited use.
Source: OpenAI announcement “ChatGPT Ads expands across Europe”, 18 August 2026, and press reports of 25 August 2026.
