The short answer
Using AI with client data can be done safely, but not by pasting it into a free consumer chatbot. Use a business arrangement where you know where data is held, a data processing agreement is in place, client data is not used to train models, access is limited and logged, and data can be deleted on request. Keep a person approving anything that reaches a client. This guide is general information, not legal advice: check the rules that apply to your firm.
- Known data location
- Data processing agreement
- No training on your data
- Least-access, logged
- Deletion on request
- Personal AI accounts
- Keys and passwords in chat
- Unchecked output to clients
- Unknown sub-processors
What can go wrong
Most problems are not exotic hacks. They are ordinary: a team member pastes a client file into a personal AI account, a provider keeps data longer than expected, a password ends up in a chat thread, or a draft goes to a client before anyone has checked it. A sensible setup makes each of these far less likely.
What a safe setup looks like
| Area | What good looks like |
|---|---|
| Location | Data held in an agreed region |
| Agreement | A signed data processing agreement |
| Training | Client data not used to train models |
| Access | Least access per person and tool, logged |
| Approval | A person approves anything client-facing |
| Deletion | Data deleted or returned when you ask, in line with your own retention duties |
UK and UAE in one paragraph
In the UK, UK GDPR and the Data Protection Act 2018 apply, and regulated firms also have their regulator’s expectations to meet. In the UAE there is a federal data protection law, Federal Decree-Law No. 45 of 2021 (the PDPL), while the DIFC and ADGM financial free zones have their own regimes. This is general information, not legal advice: confirm what applies to your firm with your compliance team or adviser.
Step by step
- 01
List the client data you would use
Names and contact details, financial information, documents, conversations. The more sensitive the data, the stricter the setup needs to be.
- 02
Set a rule for consumer tools
Make it clear to the team: no client data in free or personal AI accounts. Many data problems with AI start here.
- 03
Choose where data is held
Decide the region your data should stay in, such as the UK, the EU or a specific country, and confirm your provider can keep it there.
- 04
Put a data processing agreement in place
Any provider handling client data for you should sign one. It sets out what they can do with the data, security, sub-processors and deletion.
- 05
Confirm no training on your data
Get it in writing that your inputs and client data are not used to train AI models, including by the provider’s own suppliers.
- 06
Limit and log access
Give each tool and person the least access that works. Keep credentials in a password manager, not in chats or emails. Make sure activity is logged.
- 07
Keep a person approving
Nothing AI-drafted should reach a client without a named person checking it. Log changes, so you can show who approved what.
- 08
Plan for deletion and incidents
Know how a provider will delete or return data when you ask, in line with your own record-keeping duties, how long it keeps data, and how it will tell you about an incident.
Common mistakes
- Data locationUK region confirmed in writingConfirmed
- Training on inputsNot answered yetAsk
- Data processing agreementDraft receivedIn review
- Staff using personal accountsPolicy not writtenGap
- Staff pasting client documents into personal AI accounts because it is quicker.
- No data processing agreement with the AI provider.
- Assuming “enterprise” means data is not used for training, without checking.
- Sharing passwords and API keys in chat messages.
- Letting AI output reach clients without a person approving it.
Questions
Is it safe to put client data into ChatGPT?
Not into a free or personal account. Business arrangements exist where data handling is agreed, a data processing agreement is in place and data is not used for training. Check what your firm has agreed and what its policies allow.
Does UK GDPR allow AI to process client data?
UK GDPR does not ban it, but the usual rules apply: a lawful basis, transparency, security, data minimisation and appropriate agreements with processors. A data protection impact assessment is often expected for new uses of personal data like this. Take advice for your circumstances; this guide is not legal advice.
Do UAE data protection laws allow AI to process client data?
They do not ban it. The UAE has a federal data protection law, Federal Decree-Law No. 45 of 2021 (the PDPL), and the DIFC and ADGM financial free zones have their own data protection regimes. Which applies depends on where your firm is established and operates. Check with your adviser.
What should I ask an AI provider about data?
Where data is held, whether it is used for training, who can access it, which sub-processors they use, how long it is kept, how to delete it, and how they report incidents. Ask for a data processing agreement.