FamilyAtlasDo this first
- Stop adding personal information; identify the actual service/account.
- Secure exposed passwords or codes through the affected account provider.
- Contact the issuer or school administrator for relevant exposure.
- Review conversation, file and shared-link controls separately.
- Use safeguarding help for threats or sexual exploitation.
First identify the service, account type and information shared. A first name in a private prompt is not the same incident as a password, verification code, bank detail, precise location or a public shared link. Stop further sharing and address the actual exposure; do not assume every chatbot conversation was published online.
If credentials were entered, protect the real account through its official security process rather than waiting for chat deletion. For payment or identity exposure, ask the relevant provider what action is needed. Review the chatbot’s privacy, history, memory, training and sharing controls separately. Deleting a conversation or turning off model training does not necessarily delete every file, recipient copy or retained backend record. School-managed tools can have different terms and an administrator’s response process. FamilyAtlas does not read private chatbot conversations or guarantee removal of information already shared.
OpenAI: Data controls in ChatGPT · OpenAI: Chat and file retention in ChatGPT · Google: Gemini Apps Privacy Hub
The parent concern
Parents need to know what was shared and with which service before deciding what to secure, delete or report. Different data calls for different action.
Triage the information without copying it elsewhere
Ask what was entered, uploaded or connected: name; school/contact details; precise whereabouts; photos; medical details; passwords/codes; payment or identity documents. Check whether a link was shared or an external tool was involved. This practical triage is not a claim that every disclosure causes identity theft. Do not paste the sensitive material into another chatbot to ask whether it is safe.
Secure exposed credentials and payments first
Treat a disclosed password or code as an account-security incident and use the affected provider’s guidance. For ongoing fraud or exposed payment details, contact the issuer through a trusted route. For school credentials/data, notify the administrator. Do not wait for a chatbot deletion request before containing a usable credential exposure.
FTC: How To Recognize and Avoid Phishing Scams
Use the credential and phishing containment checklist · Recover a compromised Google Account · Secure an affected Apple Account
ChatGPT: training, history, files and sharing are different
For eligible personal accounts, Settings > Data controls > Improve the model for everyone governs future training use; turning it off does not remove existing chats. Chat deletion is scheduled for backend deletion within 30 days with stated exceptions. Files saved separately in Library can require separate deletion. Shared links need review, and deleting a link does not remove copies recipients already saved. Managed workspace controls differ.
OpenAI: Data controls in ChatGPT · OpenAI: Chat and file retention in ChatGPT · OpenAI: Sharing conversations and scheduled tasks in ChatGPT
Gemini and other tools: use the actual privacy notice
Gemini’s Privacy Hub distinguishes consumer use from work/school accounts. Read its activity, review and retention terms; deleting visible activity is not a universal promise that every reviewed record is gone. For any chatbot, identify the actual provider and ask about deletion, files, connected apps and sharing. Do not assume ChatGPT settings or deadlines apply to it.
If the material creates a safety risk
If intimate images, threats, coercion or identifying details create danger, use child-protection help rather than only a settings change. Do not download or forward sexual images of minors to prove the disclosure. Use the dedicated response guide and appropriate local emergency help when necessary. Afterwards, agree on what never goes into prompts and who can help review a confusing tool.
NCMEC: Generative AI and Child Safety · NCMEC: Take It Down: frequently asked questions
Use the safeguarding path for sexualized AI content · Set practical boundaries for ChatGPT and schoolwork
What you can do right now
- Identify service, account and data shared.
- Stop further disclosure and avoid new copies.
- Secure exposed credentials/payments promptly.
- Review chats, files, shared links and privacy controls separately.
- Contact the school/provider or safeguarding authority as appropriate.
A way to start the conversation
“Thank you for showing me. We need to know which tool and what information, not copy it into another chat. We will protect the affected account first and review privacy together.”
A suggested script to adapt, not a clinical intervention.
What technology cannot solve
Deletion, training opt-out, memory and sharing have different effects and provider-specific exceptions. FamilyAtlas cannot inspect private AI chats, retract every copy or guarantee backend erasure.
Questions parents ask
Does turning off training delete the chat?
No. Review history/deletion, files, memory and shared links separately according to the provider.
Was the whole conversation automatically public?
Do not assume that. Check the actual service, account type, sharing links and connected tools.
Does chat deletion remove every uploaded file?
Not necessarily. OpenAI documents separately saved Library files; other providers have their own retention rules.
Sources and review
Provider documentation explains settings and eligibility. Professional guidance provides context, while our suggested family strategies are labeled separately. FamilyAtlas claims were checked against implementation, with release limits stated above.
- OpenAI — Data controls in ChatGPT fact · checked 2026-09-30
- OpenAI — Chat and file retention in ChatGPT fact · checked 2026-09-30
- Google — Gemini Apps Privacy Hub fact · checked 2026-09-30
- FTC — How To Recognize and Avoid Phishing Scams safety · checked 2026-09-30
- OpenAI — Sharing conversations and scheduled tasks in ChatGPT fact · checked 2026-09-30
- NCMEC — Generative AI and Child Safety safety · checked 2026-09-30
- NCMEC — Take It Down: frequently asked questions safety · checked 2026-09-30
Next source review: 2026-12-30. Menus, eligibility and rollouts can change; follow the linked provider instructions if your screen differs.
Educational guidance, not medical advice. The cited professional guidance is separate from our suggested family strategies. This article has not had independent clinical review; seek qualified help for persistent concerns or urgent safety risks.
Written by FamilyAtlas; product descriptions are not independent product recommendations. No claim of medical or psychological review is made.