top of page

Is it safe to upload company data to ChatGPT and other AI tools?

  • 20 hours ago
  • 7 min read

We get this question in almost every workshop we run.


Someone puts their hand up halfway through and asks it. Usually about a client contract, a price list, or a CV.


So we finally wrote the answer down.


Short version: on a business subscription, yes for most everyday work. ChatGPT, Claude, Gemini and Microsoft 365 Copilot do not use your data to train their models by default, and they carry the same certifications as the cloud tools you already trust. The risk is not the tool. It is the account your team actually uses, and the apps you connect to it.


Below is what each provider commits to, what they do not, and the one thing that should never go into a chat window.


1. Do AI companies train their models on your data?

Not on business subscriptions. ChatGPT, Claude, Gemini and Microsoft 365 Copilot all state that customer inputs and outputs are excluded from model training by default on their business and enterprise plans. Personal and free accounts are different. There, your data may be used, and the controls vary by product. The account tier decides this, not the tool.


This is stated in the providers' own privacy documentation and commercial terms.

ChatGPT, Claude, Gemini and Copilot all exclude business and enterprise data from model training by default. On personal plans, data may be used and the controls vary. Sources: OpenAI DPA, Anthropic DPA, Google DPA, Microsoft DPA.


A DPA is a contract that explains how a provider must handle and protect personal data on behalf of your company.


Exceptions apply if your organization explicitly chooses to share data. For example by submitting feedback to the provider, or joining an opt-in program.


Here is the part most companies miss. A team on a business plan is covered. The same person, on a Sunday evening, using a free personal account on their phone, is not. Same tool. Completely different terms.


Get the account tier right and most of this problem disappears.


2. How do business AI subscriptions protect company data?

Business subscriptions add protections that personal accounts do not have: a Data Processing Agreement, ISO 27001 and SOC 2 Type II certification, GDPR terms, and in several cases ISO 27701 and ISO 42001. Admin controls over access and retention come with them. This is the same class of protection you already accept when you keep files in Google Drive or OneDrive.


The table below compares selected audits, certifications and privacy protections across the four tools.

SOC 2 Type II, DPA and GDPR support, and ISO 42001 are publicly confirmed for all four providers on both business and enterprise plans. ISO 27001 is confirmed for Claude, Gemini and Copilot on both tiers, and for ChatGPT Enterprise. For ChatGPT Business it is not clearly confirmed. ISO 27701 is confirmed for Gemini and Copilot, and is not publicly listed for Claude. Verified August 2026 against the providers' public trust centres.


What do these labels mean?

  • ISO 27001: An international standard for managing information security. It shows the provider has a structured system for identifying and managing security risks.

  • SOC 2 Type II: An independent audit of how selected security and privacy controls actually operated over a period of time.

  • GDPR: The EU privacy law governing how personal data is processed. Providers can offer DPAs and controls to support compliance. Your company still remains responsible for how it uses personal data.

  • ISO 27701: A privacy extension to ISO 27001. It adds a formal system for managing and protecting personal data.

  • ISO 42001: An international standard for AI management systems. It covers AI-related responsibilities, risks, controls and oversight.


All four providers offer strong enterprise-level protections. They are also used by large, heavily regulated organizations.

Examples from the providers' own published customer stories include Morgan Stanley, PwC, BBVA and Deutsche Telekom on ChatGPT; BCG, GitLab, Notion and LexisNexis on Claude; Uber, Deloitte, DocuSign and Robinhood on Gemini; and PwC, University of Oxford, DLA Piper, KPMG and Vodafone on Copilot. This shows professional adoption, not what data each organization uploads.


If banks and law firms have signed off on these terms, the terms are not the weak link in your setup.


3. What risks remain when you upload company data to AI tools?

Five risks survive even on a business plan: an external attacker breaching the provider, an insider at the provider misusing access, someone inside your own company reaching an account or chat they should not, disclosure during litigation or a lawful investigation, and connected apps or APIs forwarding data elsewhere. Four of those five are not specific to AI at all.


Business AI subscriptions provide strong protection. But as with any cloud service, some risks remain:

  • An external attacker could breach the provider and access customer data

  • An employee or contractor at the provider could misuse their authorized access

  • Someone inside your company could gain unauthorized access to an account, chat, or uploaded file

  • Chats and files could need to be disclosed during litigation, an investigation, or another valid legal process

  • Connected apps, agents, or APIs could send the data to other providers


Most of these are not unique to AI. The same risks exist when company data sits in Google Drive, OneDrive, or any other cloud service.


That is worth saying out loud. AI tools are often held to a standard that the rest of the company's software stack has never been held to.


Google says Workspace with Gemini receives the same data protection and security standards as other Google Workspace services.


Microsoft says Microsoft 365 Copilot operates under its existing Microsoft 365 privacy, security, and compliance commitments.


Judge AI tools by the same standard you already apply to your cloud storage. Not a stricter one, and not a looser one.


4. What changes when you connect other applications to an AI tool?

Every connection adds another company that touches your data. If ChatGPT sends customer records to your CRM through an app or API, the data is no longer protected by ChatGPT's controls alone. The integration provider and the CRM matter just as much. Risk follows the full data path, so the more services the data crosses, the more security boundaries you need to check.


When you connect an AI tool to another application, your data may be handled by more than one provider.


This does not automatically make the setup unsafe. But you also need to trust how each connected service protects and uses the data.

Company data moves from your business AI tool, which has enterprise controls and no training on your data by default, into a connected app or API with separate terms, and then into another system. Each step adds a security boundary to review.


For example, if ChatGPT sends customer data to a CRM through an app or API, the data is protected not only by ChatGPT's controls. The app, the integration provider, and the CRM also matter.


In other words, the risk depends on the full data path.


This is the part that changes fastest right now. Connectors, agents and MCP integrations are being switched on in companies every week, often without anyone checking who is now in the chain.


Before you switch on a connector, write down which providers now touch the data. If you cannot name them, do not switch it on yet.


5. What should you never upload to AI?

One thing is absolute on any plan: credentials. Passwords, API keys, access tokens. Beyond that, what you may upload depends on your sector, your contracts and your own policy, so it is a question for your legal team or DPO. The practical failure we see is almost never the provider. It is a rushed employee on a personal account.


Everything else is a policy question, not a technical one.


Write it down once, on one page, and your team stops guessing.


A one-page AI guideline solves more of this than any procurement process.


Conclusion

So, is it safe to upload company data to ChatGPT and other AI tools?


For most everyday business uses, yes, if you use a business subscription. The business products from ChatGPT, Claude, Gemini and Microsoft 365 Copilot offer strong security and privacy protections. Customer data is generally not used to train the providers' models by default, and major organizations already use these tools in professional settings.


No cloud service is risk-free. But three things get you most of the way there:

  1. Put your team on business accounts, not personal ones

  2. Know which providers sit in the chain before you connect anything

  3. Keep credentials out of the chat window


With those in place, companies can use AI confidently to work faster and improve the quality of their work.


We hope this helps you make a more informed decision. If you have a different perspective, or think we have missed something, leave a comment or send us an email. We update this article as the providers change their terms.


And if you need support implementing AI solutions in your business, that is what we are experts in. Please contact us.


FAQ

Is ChatGPT safe for company data?

For most everyday business uses, yes, if you use a business subscription. These accounts provide stronger privacy and security protections than personal accounts.

OpenAI states that data from its business products is not used to train its models by default. Personal accounts may work differently depending on their settings.

They share most cloud security risks. The difference is that AI tools actively process the information and may pass it to connected services.

Credentials, always: passwords, API keys, access tokens. Beyond that it depends on your sector and contracts, so agree it internally and write it down.

Only if your company has approved the tool and that specific type of information. Use a business account and share only the data needed for the task.

Yes. Each connected application adds another provider that may receive or process the data. You need to trust every provider in the chain, not only the AI tool.

No. The AI Act regulates how AI systems are built and disclosed. Data protection when you upload a file is governed by the GDPR.



Comments

bottom of page