Before launching an AI project, check five things: what data goes into the tool, where it is processed, what the supplier does with it, what you must tell users, and the risk level of the use. This article gives the set of questions. It does not replace legal advice.
1. What data goes into the tool?
List what the AI will read or produce: names, emails, customer messages, contracts, employee data. The more personal or sensitive the data, the more precautions increase. Where possible, give the tool only what is strictly necessary, and remove what is not useful for the task.
2. Where is the data processed?
Ask the supplier where the data is hosted and processed, and whether it leaves the European Union. Transfers outside the Union require specific safeguards. European hosting simplifies the file.
3. What does the supplier do with your data?
- Is it kept? For how long?
- Is it used to train its models? Can you opt out?
- Is there a personal data processing agreement?
- Who can access it at the supplier?
If you cannot find these answers, do not entrust personal data to the tool.
4. What must users know?
Inform the people concerned that their data is processed and for what purpose. For a chatbot or an assistant, state clearly that it is an AI: European AI regulation provides for this transparency for this type of tool. Also plan how a person can request access to their data or its deletion.
5. What is the risk level of the use?
European regulation distinguishes uses according to their risk. An assistant that summarises documents or answers frequently asked questions is not comparable to a tool that helps decide on a hire, a loan or access to a service. The greater the impact on people, the more the requirements increase. When in doubt, seek advice before launching.
The checklist to keep
- I listed the data processed and removed the superfluous.
- I know where it is processed and how long it is kept.
- I checked the contract and the supplier’s use of the data.
- Users know they are talking to an AI and how to exercise their rights.
- A person reviews high-impact decisions.
- Someone is responsible for follow-up after launch.
Getting support
Our AI transformation consulting builds compliance in from the diagnostic: use cases, data, tools, safeguards. We then build suitable AI applications and agents, and offer AI training for your teams. To start, describe your project to us.
