Being open about AI
Be clear about AI use when the law or your organisation asks for it, and avoid misleading people. That does not mean putting “written with AI” under every email. It is about the situations where it makes a difference.
- check_circle When you say that AI was used
- check_circle What customers need to know when they talk to a chatbot
- check_circle How to use meeting recordings and note-taking tools correctly
When do you say that AI helped write something?
- A customer chats with a chatbot on the website: make it clear at the first contact at the latest that it is AI, unless this is already obvious to the user. Do not wait until they ask.
- You publish images, video or voices that were created or edited by AI and look real: as a practical rule, make this recognisable as AI. The legal disclosure obligation for deepfakes concerns material that could wrongly appear real or authentic, for example images of existing people or events.
- A customer or partner asks you directly whether AI was used on their file: you answer honestly.
For an ordinary email that you wrote with the help of AI and checked yourself, there is no general obligation to label it as AI. Do follow the agreements within your organisation and with your customer.
For AI text published to inform the public about matters of public interest, there is a specific transparency rule. Among other things, it has an exception where there is human review or editorial control and editorial responsibility. Have such publications assessed beforehand in line with your communication procedures.
Recording and transcribing meetings
Recording, transcribing and summarising are different functions. Even without a stored audio recording, an AI tool can process personal data. So:
- Inform people beforehand, including external participants: which function are you using, what for, who gets access and how long is the data kept? Follow the approved procedure for this.
- Ask whether it is okay and take objections seriously. If someone does not agree, do not switch the function on, or stop it. Take notes yourself and have any follow-up step assessed under the internal procedure.
- Use only the permitted tool and function. No separate note-taking apps or bots that you add to a meeting yourself.
Your organisation has to determine the legal basis and the other legal conditions beforehand. Consent is not the required GDPR legal basis in every situation; especially at work, it cannot always be given freely. Quickly asking whether everyone agrees does not replace that assessment.
An AI report says who said what, including about other people. So share it only with those who need it, just like any other document containing personal data.
A note-taking bot in the meeting
What do you do?
Inform beforehand and follow the approved procedure. Explain how the AI function processes what is said and how access and retention periods are arranged. That is needed even without a stored audio recording. If someone objects, you do not switch the function on and you take notes yourself. The legal basis has been assessed beforehand by the organisation; an informal agreement does not replace that.
What you take away from module 5
- bolt Make it clear at the first chatbot contact at the latest that it is AI, unless that is already obvious.
- bolt As a practical rule, make realistic-looking AI material recognisable as AI and have public communication assessed in line with the agreed procedures.
- bolt If someone asks whether AI was used, you answer honestly.
- bolt Inform people beforehand about AI recordings, transcripts and summaries. Follow the approved procedure and stop if someone objects.
Module 5 complete 🎉
You now know when and how to be open about AI. In module 6 you will see how scammers use AI against you, and how to protect yourself.
On your way to the “Using AI safely” practice quiz
Complete all 8 modules and try the free practice quiz. The official exam with a certificate is in the GDPRWise app.