Data privacy traffic lights
About this app
Explore which administrative information belongs in a prompt, needs anonymisation or an approved tool, or must stay out of public AI. Includes explanations and a self-test; no AI calls.
The idea
Typing text into an AI tool passes it on to another system. Before every prompt, public bodies and companies therefore have to ask whether the information may go there. Personal data fall under the GDPR, and health and social data belong to the special categories in Article 9 GDPR. German public administration adds social-data secrecy (§ 35 SGB I), tax secrecy (§ 30 AO) and official confidentiality. General questions without personal data, on the other hand, are usually fine. A rule of thumb: when in doubt, anonymise or leave it out.
What you see
A 3D traffic light and a list of typical cases from everyday administrative work, grouped as green (safe), amber (only anonymised or only in an officially approved tool) and red (never in public or unapproved AI). Examples include a citizen’s name and address, a tax ID, an anonymised case history, a screenshot or a general wording question. For each case the app gives a reason and a recommended action.
Things to try
Pick a case and read why the light shows that colour. In the self-test you get a case and decide between green, amber and red yourself; a counter tracks how many you classified correctly. Pay attention to the amber cases: even without real names, details, metadata or small case numbers can make a person identifiable. The ratings are a teaching aid; in a specific case the data protection officer decides.
Frequently asked questions
- Can I enter personal data into ChatGPT or other AI tools?
- Under the GDPR, personal data may only be processed with a legal basis and with the transfer to the provider properly secured. As a rule they therefore do not belong in publicly available, unapproved AI services. Organisations usually handle this through approved tools and internal policies.
- Is removing names enough before using AI?
- Removing names lowers the risk but does not always eliminate it. Case numbers, places, rare details, image edges or metadata can still make a person identifiable. Anonymised content should therefore be checked first and ideally used only in approved tools.
- Which data are specially protected?
- Article 9 GDPR lists special categories of personal data, including health data, genetic and biometric data, religious beliefs and political opinions. In German public administration, social-data secrecy and tax secrecy protect certain data in addition.
- What can I do in this app?
- You pick typical cases from a list or at random and see the green, amber or red rating with its reason and recommendation. In the self-test you classify cases yourself and get immediate feedback. The app makes no AI requests.
Subject: Machine Learning
More from Machine Learning
- AdaBoost
- Anatomy of a prompt
- Decision Tree
- Ensemble: Bagging vs. Boosting
- Finding faces: masks versus learning from examples
- k-Means
- k-Nearest Neighbors (kNN)
- AI theatre
A VisuApp by heyprof: interactive, free in your browser, no sign-up. What is a VisuApp? · All apps