ChatGPT Safety Settings

ChatGPT safety settings help parents guide a teenager’s use of OpenAI’s conversational AI service. It can explain ideas, draft text and work with voice, images or files where available. Young people may use it for study, creative projects and everyday questions. However, a confident answer can still contain mistakes.

For parents, begin with the age rules and account type, then review the controls that apply to your device. This guide separates protective settings from family advice. In addition, check permissions again after a major update or a change of account.

Age Rating and Child Safety Warning

ChatGPT is not intended for children under 13. Users must meet the minimum age required in their country, and anyone under 18 needs a parent or guardian’s permission. App store classifications describe the app rather than overriding these account rules. Use a genuine birth date and a separate eligible teen account, rather than sharing an adult login. Read the provider’s age guidance.

ChatGPT for Teens adds protections when the account is identified as belonging to a minor. Age prediction can make mistakes. Where verification is offered, use the instructions inside the official account and read the provider’s privacy notice before submitting an image or identity document. An age check does not establish that every answer is suitable. Review the relevant account guidance.

Restrictions Available

Link the Parent and Teen Accounts

Linking gives an adult access to selected controls without access to the teen’s transcripts.

• Parent: Open the profile menu, then Settings and Parental controls.
• Next, select Add family member, choose Use phone or Use email, enter the teen’s details and select Send.
• Teen: Open Settings then Parental controls and select Accept invite.
• Finally, parent: Select the teen under Family members to manage controls.

Why it matters: However, a teen can unlink the accounts. In addition, parents receive notice, but these controls cannot cover another account.

Review Every Linked Teen Feature

The linked panel includes content, learning, privacy and tool controls.

• Open Settings, Parental controls and the teen’s name.
• Next, keep Reduce sensitive content on. Review Improve the model for everyone and Reference saved memories.
• Finally, review Voice Mode, Image generation, Study mode, Codex network access and Cloud browser.

Why it matters: However, turning off code network access does not block search, connected apps or the separate browser. In addition, feature changes may require a new conversation.

Set Quiet Hours and Study Hours

These schedules have different purposes. Quiet hours restrict access, while study hours change how eligible new chats begin.

• Parent: Open Settings, Parental controls and the teen’s name.
• Next, enable Quiet hours and select a start and end time. Add additional blocks if needed.
• Finally, separately review Study hours and set the desired study schedule.

Why it matters: However, study hours do not block access and do not apply to GPT or Project conversations. In addition, agree routines that also cover other devices.

Review Safety Notifications

Linked accounts can receive limited safety notifications. These are not a live monitoring service. When an account moves out of the teen experience at 18, its parental connection ends and teen safety notifications stop.

• Parent: In the teen’s Parental controls, find Safety notifications.
• Next, select Manage notifications and review the available contact methods.

Why it matters: However, notifications may miss concerns and do not include transcripts. Therefore, keep ordinary conversations with your teen central to the family plan.

ChatGPT Safety Settings

Some ChatGPT safety settings depend on the plan, region or account. Use the version specific instructions below, and keep the provider’s help page open if your menu differs. A missing option should never be treated as an active protection.

ChatGPT: Turn Off Model Training

This controls use of new conversations to improve models. It is separate from chat deletion.

• Open the account menu, then Settings and Data controls.
• Next, turn off Improve the model for everyone and select Done if shown. On mobile, open the sidebar and profile first.

Why it matters: However, saved history remains. In addition, submitting response feedback can make the associated conversation available for model improvement, so avoid including private information.

Review and Remove Memory

Memory can draw on chats and other available sources. Switching it off does not erase those sources.

• Open Settings, Personalization and Memory.
• Next, review the available Manage, Saved memories or memory summary controls. Remove unwanted entries and disable memory if appropriate.
• Finally, delete the relevant chats and files separately, and disconnect any app containing the information.

Why it matters: However, a remembered detail can survive deletion of its original chat. Therefore, review all relevant sources before treating information as removed.

ChatGPT: Choose an Unpersonalized Temporary Chat

Temporary chats have their own history and personalization choices.

• Start New chat and select Temporary.
• Next, choose Unpersonalized before sending the first message.

Why it matters: However, while temporary, the chat stays out of history and does not create memories or train models. In addition, a copy may remain for 30 days. Saving it makes it a regular chat.

Protect Sign In and End Other Sessions

Additional verification helps protect chats, uploaded files and connected services.

• Open Settings and Security and login.
• Next, under Multi-factor authentication (MFA), select an available method and complete its setup. Keep recovery access secure.
• Finally, to end existing sessions, return to Security and login and select Log out of all devices.

Why it matters: However, enabling extra verification does not end existing sessions. Therefore, sign out shared computers and protect the email account used for recovery.

ChatGPT: Disconnect Apps and Review Action Permissions

Connected services can expose more information than a typed question.

• Open Settings and Plugins. Some interfaces list Apps instead.
• Next, choose the relevant app or plugin and review Connected accounts or Connection.
• Finally, open the account’s more options menu and select Disconnect. Read any warning about other linked accounts.

Why it matters: However, disconnection stops future access through that connection. In addition, it does not erase existing chats, files or memories that already contain the information.

Review Shared Links

Sharing can expose the conversation’s contents to the permitted audience.

• Open Settings, Data controls, then Shared links and Manage.
• Next, choose Delete shared link beside an unwanted item. To remove all links, open More actions and select Delete all shared links.
• Finally, before creating another link through Share, inspect its content and audience.

Why it matters: Removing a link cannot recall screenshots or copies already made. In addition, never include identifying school or family details in a shared conversation.

ChatGPT: Export Data and Delete the Account

Export first if you need a personal record. Account deletion is permanent.

• Open Settings and Data controls. Under Export data, select Export and complete the verification instructions.
• Next, to close the account, open Settings, Account, then Delete beside Delete account. Enter the requested account identifier and DELETE, then select Permanently delete my account.

Why it matters: Deletion affects access across the associated OpenAI account. Therefore, check subscriptions and downloaded copies separately before proceeding.

Cancel Renewals Through the Billing Platform

Uninstalling the app does not stop subscription charges.

• For a web purchase, open Settings, Billing and select Cancel under Cancel plan.
• Next, for an Apple purchase, open device Settings, your name and Subscriptions. Select ChatGPT and Cancel Subscription.
• Finally, for Google Play, open Subscriptions, select ChatGPT and Cancel subscription. Confirm the final status.

Why it matters: Cancellation stops renewal rather than automatically refunding past charges. Therefore, cancel a store subscription before deleting the account.

ChatGPT Safety Settings for App Actions

App permissions control approvals separately from account connections.

• Open Settings, then Plugins or Apps. Open Permissions where available.
• Choose Always ask to review access and changes. For a specific connection, select its app, open the Connected accounts menu, then Settings and Permissions.
• Select Save if shown. Read approval prompts before choosing Allow once or Deny.

Why it matters: Remembered approvals can reduce later prompts. However, they do not replace provider permissions. Managed accounts may apply different rules.

Delete Saved and Archived Chats

Archiving keeps a conversation in the account.

• Open chat history, select its … menu, choose Delete and confirm.
• Open Settings, Data controls, then Archived chats or View archived chats. Select Manage if shown and delete unwanted items.
• To remove all chats, use Settings, Data controls and Delete all chats. Read the confirmation.

Why it matters: Deletion cannot be undone. Furthermore, Library files and separate memories need their own review.

Stop Scheduled Work and Notifications

Tasks can continue after you close a conversation.

• Open Scheduled. Alternatively, open Settings, Notifications, then Manage tasks.
• Open a task and use its panel to Pause or Delete it.
• Review Settings and Notifications separately for Push and Email choices.

Why it matters: Disabling notifications does not stop tasks. In addition, deleting a task’s chat pauses it but can leave its shared link available.

How to Use ChatGPT Safely

For parents, review image, microphone, camera and screen sharing before a voice session. Show your teen how to stop sharing before opening private notifications. In addition, treat instructions found in web pages or files with caution when an AI tool can act on connected accounts. A helpful looking document can contain misleading instructions.

Review ChatGPT safety settings together, then ask for explanations rather than accepting an answer you cannot follow. For example, compare a generated summary with the original material and check unfamiliar links before opening them. Keep passwords, addresses and private photographs out of prompts, even when a privacy switch is off.

Essential Safety Advice and Current Concerns

AI can produce persuasive errors, imitate people and generate harmful material. A friendly reply can also encourage someone to reveal more than they intended. Therefore, discuss what your family considers private and how to leave an uncomfortable interaction without shame.

Never use an assistant’s reassurance as proof that a person, investment or link is genuine. If another user pressures a young person to send images, money or passwords, stop the interaction and seek help. In addition, agree a human contact for problems that feel difficult to explain.

Report a Harmful Response

Reporting flags a reply for review. It does not guarantee an immediate response.

• Select Thumbs down beneath the reply and choose the relevant reporting reason.
• Use the official content reporting form for content that may breach rules or law.

Why it matters: Stop a troubling conversation and seek human help as needed. A report should contain only the information needed to explain the issue.

Listen calmly if your child reports a worrying experience. For urgent safeguarding concerns, seek appropriate help rather than waiting for a platform reply. In immediate danger, contact emergency services on 112 or 999 in Ireland. CODA’s contact form is for education enquiries.

Platform Support and Regional Differences

In Ireland, the UK and the USA, follow the service’s current terms and the controls available to the actual account. Privacy law and a provider’s minimum account age are separate matters. A privacy consent rule does not give a child permission to ignore a stricter service age limit.

For privacy concerns, start with the provider’s support or privacy request route. Families in Ireland can consult the Data Protection Commission’s parent guidance. UK families can use ICO guidance, while US families can consult FTC children’s privacy guidance. These resources explain rights and reporting, rather than certifying any app as safe.

ChatGPT Safety Settings Checklist

• Confirm the service age rules and use the correct account type.
• Review saved activity, memory, sharing and connected services.
• Protect sign in, recovery access and the devices used.
• Check purchases and renewal dates before accepting a trial.
• Practise how to stop a chat, report a problem and ask a trusted person for help.

ChatGPT Safety Settings FAQ

Do these settings make ChatGPT safe for every child?

No. Begin with the provider’s age rule and suitability for the individual child. Controls reduce particular risks, while content and the way someone uses an assistant can still cause problems.

Does turning off training delete conversations?

No. Training choices, history deletion and memory are separate controls. Review each one and check information already shared with other services or people.

Should a parent share an adult login?

Avoid sharing an adult account. It can expose private history, payment details and connected services, while bypassing protections attached to a correctly identified teen account.

Related CODA Guides

Return to AI Assistants or AI Safety Settings. For wider family guidance, explore Resources for Parents.

References

Account age requirements: official instructions for the account rule or control above.

Account eligibility and protections: official instructions for the account rule or control above.

Link the Parent and Teen Accounts: official instructions for the account rule or control above.

Set Quiet Hours and Study Hours: official instructions for the account rule or control above.

Review Safety Notifications: official instructions for the account rule or control above.

ChatGPT: Turn Off Model Training: official instructions for the account rule or control above.

Review and Remove Memory: official instructions for the account rule or control above.

ChatGPT: Choose an Unpersonalized Temporary Chat: official instructions for the account rule or control above.

Protect Sign In and End Other Sessions: official instructions for the account rule or control above.

ChatGPT: Disconnect Apps and Review Action Permissions: official instructions for the account rule or control above.

Review Shared Links: official instructions for the account rule or control above.

ChatGPT: Export Data and Delete the Account: official instructions for the account rule or control above.

Cancel Renewals Through the Billing Platform: official instructions for the account rule or control above.

Report a Harmful Response: official instructions for the account rule or control above.

DPC parent privacy guides, ICO public guidance and FTC children’s privacy information. Regional rights and support resources.

• App action permissions Approval choices and connection limits.
• Deleting and archiving chats History controls and separate file retention.
• Scheduled task controls Pausing tasks and managing notifications.
• Changes at age 18 Ending the teen experience.

About Children of the Digital Age

Children of the Digital Age provides online safety education and training for families, schools and organisations. Our work connects practical safety guidance with cybersecurity and an understanding of online behaviour.

Send a message

For education about ChatGPT safety settings, contact us to discuss parent sessions or school presentations. Tell us which service your family uses and the topic you would like to understand.

Last updated: 15 October 2026