Guide for consultants · AI
Can our team put client data into ChatGPT?
The short answer: personal or confidential client data does not belong in a personal ChatGPT account. In a business or enterprise plan with a data processing agreement it can be permissible, provided you clarify a few points first. This guide explains the difference between personal accounts and business plans, what to discuss with your data protection officer and what a simple AI policy for your team looks like. It is written for companies in the EU and does not replace legal advice. It helps you ask the right questions.
What client data means here
At a consultancy or professional services firm, two kinds of client data quickly end up in AI tools:
• Personal data: names, email addresses, phone numbers, but also meeting notes about specific people. The GDPR applies to them.
• Confidential business information: proposals, figures, strategies and internal documents of a client. Confidentiality clauses in your contract with the client usually apply to them.
Both can be in the same email. Anyone who pastes a client email into an AI tool to draft a reply usually shares both.
Personal accounts: Free, Plus and Pro
For personal ChatGPT accounts, OpenAI says that sharing data for model training is on by default. You can turn it off under “Settings”, “Data controls” with the option “Improve the model for everyone”. New conversations are then no longer used for training. Temporary Chats are not used for training either, but may be kept for up to 30 days for safety purposes.
That does not solve the core problem. Personal accounts come without a data processing agreement with your company. If employees enter client data there, the contractual basis for sharing it is usually missing. That is a risk for your company, not only for the individual.
Business offerings: ChatGPT Business, Enterprise and Microsoft Copilot
Business offerings work differently:
• ChatGPT Business and ChatGPT Enterprise: OpenAI says it does not train its models on business data by default. For ChatGPT Business, ChatGPT Enterprise and the API, OpenAI offers a data processing addendum. Check your contract documents to see whether it already applies to your agreement or needs to be signed separately. OpenAI provides a form for this.
• Microsoft Copilot Chat and the Copilot license (formerly Microsoft 365 Copilot): With a work or school account, Microsoft says its Data Protection Addendum and Product Terms apply, with Microsoft acting as a data processor. Prompts, responses and data from Microsoft 365 are not used to train foundation models. Microsoft names one exception itself: web searches that Copilot sends to Bing are not covered by this addendum or by the EU Data Boundary. For those searches, Microsoft acts as an independent data controller.
So a business plan is usually a prerequisite, not a green light. What else to clarify is in the next section.
What to clarify with your data protection officer or privacy advisor
These questions should have answers before client data regularly goes into an AI tool:
1. Legal basis: On what legal basis do you process the data for this purpose?
2. Agreement with the provider: Is the data processing agreement in place, and where is the data processed?
3. Agreements with your clients: Do your confidentiality clauses allow you to share client information with such a service? Some clients explicitly rule it out.
4. Special obligations: Are you or your clients bound by professional secrecy, for example in tax advice, auditing or healthcare? Stricter requirements then apply.
5. Transparency: Do your privacy notices or your record of processing activities need an update?
6. Storage and access: How long are conversations kept, and who can see them?
Why a ban alone is not enough
Many companies say “no AI with client data” and leave it at that. The work remains, though. According to a representative survey by the German Economic Institute (IW) from March and April 2026, just under three in ten employees who use AI at work use tools their employer has not provided. The institute therefore recommends relying not on bans alone but on privacy-compliant tools, skills development and clear usage rules.
For you, that means providing a business tool and clearly defining what may go into it. There is also a legal obligation: Article 4 of the EU AI Act requires companies that deploy AI to take measures to support their staff's AI literacy. A short policy with a briefing for the team is a simple start.
A one-page AI policy
A policy your team actually reads fits on one page. It answers six questions:
1. Which tool is approved, with which account?
2. Which data may go in, for example your own text, anonymized notes, public information?
3. Which data never goes in, for example health data, bank details, passwords, documents from clients who have explicitly ruled it out?
4. How to anonymize when you cannot do without client data, for example “Client A” instead of the company name? Initials or placeholders are not enough if the context still shows who is meant.
5. Who reviews results before they go to a client?
6. Whom to ask when in doubt?
Where aenuma fits
aenuma implements agreed workflows for recurring tasks such as reply drafts, meeting preparation or proposal drafts in the tools your company has already approved. Before implementation, we agree with you which information the workflow uses, which steps a person approves and how exceptions are handled. Sensitive information can be anonymized for the initial discussion.
Legal questions such as the legal basis, your client agreements or professional secrecy are for your data protection officer or legal counsel. Your requirements on these points feed into the implementation.
Frequently asked questions
Is it enough to turn off training in a personal account?
No. Turning it off prevents new conversations from being used for training. It does not replace a data processing agreement between your company and the provider. For personal or confidential client data, a business plan is usually the prerequisite.
Is a Temporary Chat safe enough for client data?
According to OpenAI, Temporary Chats are not used for training but may be kept for up to 30 days. A personal account still has no data processing agreement. That makes it no solution for client data.
Does all this apply to Microsoft Copilot too?
Copilot Chat with a work account is covered by Microsoft's enterprise data protection terms, even without an additional Copilot license. The free Copilot app with a personal Microsoft account is something else. Check which account your team is actually signed in with. A comparison of the options is in the guide Copilot, ChatGPT Business or a fixed AI workflow.
Sources
Data Controls FAQ (OpenAI Help Center)
Enterprise privacy at OpenAI (OpenAI)
Enterprise data protection in Microsoft Copilot and Microsoft Copilot Chat (Microsoft Learn)
Schatten-KI: Beschäftigte sind schneller als ihre Betriebe (IW-Kurzbericht No. 63/2026, German Economic Institute, in German)
Regulation (EU) 2024/1689 on artificial intelligence (EUR-Lex, original text, Article 4 since amended by the 2026 Digital Omnibus)
By Joshua Schütt, founder of aenuma. Last updated: October 2026.
How does your team handle AI and client data today?
If you want to prepare a recurring task with AI without client data ending up in personal accounts, we are happy to discuss your use case. Not sure where to start? The Time Sink Check shows which task makes a good first workflow.
Book a call