Direct questions

19. How do I protect my child’s data during AI mentorship?

Map where information goes, minimize what is submitted and verify the settings on the actual account. Do not equate “not used for model training” with “not stored” or “not accessible to anyone else.”

OpenAI’s current help documentation distinguishes model-training settings, saved chats and memory. Turning off model improvement does not delete saved conversations; feedback submissions can also have separate implications for training use. (help.openai.com)

ChatGPT parental controls allow management of selected settings, but do not give parents access to a teen’s conversations or chat history. Either party can unlink the accounts, with notification to the parent when the teen unlinks. (help.openai.com)

Our recommended household rules:

  • Use synthetic or appropriately cleared project data.
  • Do not upload confidential employer material, family financial records, school rosters or private communications.
  • Keep household email, calendars, cloud drives and payment systems disconnected from learning prototypes.
  • Review extensions, transcription tools and file-sharing services, not only the main chatbot.
  • Separate permission to teach from permission to publish or market.
  • Recheck settings when products, accounts or workflows change.

For each tool, record data in → recipients → retention → deletion route. UNICEF’s guidance supports privacy-by-design and children’s understanding and agency; our register is a practical household implementation proposal, not UNICEF certification. (unicef.org)

Sources cited on this page

  1. Data controls in ChatGPT | OpenAI Help Center ↗help.openai.com
  2. Managing parental controls in ChatGPT | OpenAI Help Center ↗help.openai.com
  3. # Guidance on AI and Children (v3): Checklist ↗unicef.org