3. What happens with your data?
- Pflege mit KI
- Nov 5
- 3 min read
Updated: Nov 15
Family caregivers often work with sensitive information, such as details about a family member’s health or care situation. If you enter this data directly into ChatGPT, it may, depending on the version and usage, be used to improve the model. This is not always fully compliant with data protection regulations.
What counts as personal and sensitive data
Personal data includes any information that can directly or indirectly identify a person, such as name, address, date of birth, contact information, or IP address.
Sensitive personal data
Under the GDPR, these are called “special categories” of data, covering particularly protected aspects of life. According to Article 9 of the GDPR, this includes:
Health data (e.g., care level, diagnoses, medical treatments)
Ethnic origin
Political opinions
Religious or philosophical beliefs
Trade union membership
Genetic data
Biometric data used for identification
Data about sexual life or orientation
Why this distinction matters
Sensitive data is subject to stricter protection. Processing is generally prohibited unless a clearly defined exception applies, such as:
Explicit, informed consent from the individual
Processing necessary to protect vital interests if consent cannot be given
Legal obligations in health, social protection, or employment
Note: An entry like “Care level 4” counts as health data and is therefore sensitive. Handle it carefully.
How to use ChatGPT safely
Maintain anonymity
Avoid entering real names or addresses. Instead of: “Ms. Müller, 83, Berlin” Use: “the person being cared for, an elderly person in their eighties, lives in the city”
Anonymize health data Instead of: “Care level 4 with Parkinson’s” Use: “High care needs with mobility limitations”
Disable training – in both Free and Plus versions
In ChatGPT Free and Plus, the training of your inputs is enabled by default. Your inputs may be used to improve the model. You must actively opt out if you don’t want your past entries to be used.
To do this, go to: Profile → Settings → Data Controls → “Improve the model for everyone” and turn off this toggle. This will prevent your future chat content from being used for training.
Optional extra step
OpenAI develops Artificial Intelleigence, especially language models like ChatGPT. If you want extra protection, you can submit a request through the OpenAI Privacy Portal: “Do not train on my content.” From that point on, OpenAI will no longer use your content to train new models.
Use temporary chats
Temporary chats do not appear in your history, are not used for training, and are deleted after 30 days. Ideal if you want to avoid leaving any data traces.
How to create a temporary chat
Left sidebar → New Chat → Click the temporary chat icon at the top right
Advantages of temporary chats
No storage in history
Not used for training
More control over sensitive topics
Disadvantages of temporary chats
No history: You cannot access the chat later. Save manually if needed.
No multi-session context: Each chat starts fresh, which can make longer tasks cumbersome.
No personalization: Preferences, writing style, and settings are not saved.
30-day retention: Technically, even temporary chats are stored up to 30 days for abuse prevention before being deleted.
In short, temporary chats are safer but less convenient for ongoing work.
For highly personal content: work locally
Draft sensitive texts offline first (e.g., Word or another local program). Use ChatGPT only for structure suggestions or phrasing, sharing only the minimum necessary.
Conclusion
Using ChatGPT to process sensitive data carries privacy risks. By default, all inputs –even in Free or Plus – can be used to improve the model unless you opt out in settings. Temporary chats provide extra protection: they don’t appear in history and aren’t used for training.
Even without training, data can be stored for up to 30 days to prevent abuse. OpenAI is exploring stronger protections like end-to-end encryption, and legal requirements may affect data retention.
Ultimately, it’s up to each user to balance functionality and privacy. Choose the approach that gives you the best mix of security and convenience.
Ressources:
The content is for informational purposes only and does not replace legal, care, or medical advice. AI models can make mistakes. Please verify important information yourself. Learn more
