Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
To check whether an AI chatbot stores your conversations or uses them to train models, open its privacy or data-controls settings while signed in and inspect training, history, memory, temporary-chat, feedback, and deletion controls separately. Also confirm whether you are using a personal account or a work, school, or business account: the rules can differ. A training opt-out usually affects future model-improvement use; it does not necessarily erase existing chats or stop safety, security, legal, or service-related processing.
What does “stored” mean, and how is it different from training?
Storage and training are separate. A provider may retain a chat so it appears in your history, provides context or personalization, supports safety investigations, or meets legal and security needs. Training is a further use of conversation data to improve or develop models.
As an Amazon Associate I earn from qualifying purchases.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchThat distinction matters when you read a setting that says your chats are “not used for training.” It is not, by itself, a promise that the provider does not keep the conversation, that no person can review it, or that the chat is excluded from every other product or safety process. Check each relevant control and disclosure individually.
- History or activity: whether chats are saved in your account or shown in the app.
- Model improvement or training: whether conversation content may be used to improve models.
- Memory or personalization: whether information from chats can be retained to tailor later responses.
- Temporary or incognito chat: whether a particular conversation is left out of history, memory, or training—and how long it may still be retained.
- Feedback and human review: whether a rating or safety review can bring a conversation into an improvement or review workflow.
- Deletion and export: what you can remove or download, and what the provider says may remain in backups or other systems.
These categories can overlap operationally without being the same legal or technical process. For example, a chat can be retained briefly to answer your request and block abuse without being added to a model-training dataset. Conversely, a conversation already selected for improvement may be copied into a separate pipeline, so deleting the visible thread may not instantly remove every de-identified derivative.
#1 Best Overall
How can I check a chatbot’s privacy settings?
Use the provider’s live controls while signed in, because settings and labels can change and may vary by account type, product surface, region, age, plan, or administrator policy. Record what each control says rather than treating one privacy toggle as an all-purpose switch.
- Confirm the account and product. Identify whether you are signed into a personal account, a work or school workspace, an enterprise service, or an API-based product. A work account may have administrator access and different default protections from a personal account. Check the email or organization name shown in the account menu before relying on a consumer privacy article.
- Open the official privacy or data-controls page. Find the provider’s current settings page or help information, and check the labels in the account itself. The paths below identify the controls described by the providers in their 2026 disclosures; interfaces can change. Save a copy or note the wording and date if the decision is important.
- Inspect model improvement or training. Look for a toggle or choice about using chats, prompts, or conversation activity to improve models. Read whether it applies to future chats only and whether safety review, feedback, or other opt-in events are treated differently.
- Check history and activity separately. Find whether chats are saved in account history, activity, or a conversation log, and whether there is an auto-delete or retention setting. A history switch can control what you see without controlling every backend copy.
- Check memory and temporary-chat behavior. Determine whether memory is a separate feature, whether temporary chats are excluded from history or model improvement, and what retention exceptions still apply. If the service has separate audio, voice, or image-upload controls, inspect those too.
- Review feedback and deletion controls. Find out whether submitting a rating can make a chat available for review or improvement. Delete chats or memories separately when the provider provides separate controls, and read its explanation of deletion timing, backups, safety retention, and de-identified information.
- Check access by other people. For a work or school account, look for statements about administrator search, audit logs, or compliance access. A provider can exclude prompts from foundation-model training while still making logs available to your organization.
- Recheck after changes. Product labels, defaults, and policies can change. Revisit the settings after a major update, account change, or policy revision, and repeat the check if you switch from a personal to a managed account.
Where are the controls in ChatGPT, Claude, Copilot, and Gemini?
The providers describe different controls and exceptions. The table summarizes the consumer and work-account details established in their 2026 documentation; it is not a guarantee that every interface, region, or account has identical settings.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →| Service and account | Training or model-improvement control | History, retention, or deletion | Temporary use, review, and other exceptions |
|---|---|---|---|
| ChatGPT personal account | Go to Settings → Data Controls and inspect Improve the model for everyone. Turning it off prevents new conversations from being used to train ChatGPT; the conversations still appear in chat history. | Turning off model improvement does not delete history. Delete chats separately if you want to remove them from the visible history. | Temporary Chat does not appear in history, create memories, or improve models. OpenAI says Temporary Chats are retained for 30 days for safety before deletion. |
| ChatGPT Business, Enterprise, Edu, or Healthcare workspace | OpenAI says workspace content is not used to train by default. Its policy also says business products and the API are not trained on inputs or outputs by default unless an organization explicitly opts in. | Retention and access depend on the applicable product and organization controls; the cited default-training disclosure does not establish a universal deletion period. | Do not assume personal-account controls describe a workspace. Check the organization’s settings and policy. |
| Claude Free, Pro, or Max | Inspect the privacy or model-improvement choice. Anthropic says chats and coding sessions may be used to improve models if you allow it, if a conversation is flagged for safety review, or if you otherwise opt in. | Deleting a conversation removes it from chat history immediately and from backend storage within 30 days. If improvement use is enabled, de-identified data in training pipelines may be retained for up to five years. | Safety-flagged conversations are an exception to the ordinary choice about allowing improvement use. |
| Personal Microsoft Copilot | Open the profile menu → Privacy and inspect Training on conversation activity. There is a separate voice setting. Opting out excludes future conversation activity from training, but not all other uses. | Microsoft states that the default conversation-history retention is 18 months. Users can delete individual items or all history. | The opt-out does not exclude other general product-improvement, advertising, digital-safety, security, or compliance uses. |
| Work or school Microsoft Copilot Chat | Microsoft says prompts and responses are not used to train foundation models when work or school users sign in with enterprise data protection. | Prompts, Bing queries triggered by prompts, and responses are logged. | IT administrators can use search and audit tools to view the logged information. Enterprise data protection is not the same as keeping chats invisible to an employer. |
| Google Gemini | Open Gemini Apps Activity and inspect Keep Activity. With it on, activity is saved and used to provide, develop, and improve services, including training generative AI models. With it off, future chats are not used to train Google AI unless feedback is submitted. | With Keep Activity off, Google retains future chats for 72 hours to respond and protect users. Check auto-delete and the separate setting for audio or Gemini Live recordings as well. | Temporary chats are not used to train Google AI with human reviewers. A subset of chats may be reviewed by people; Google says reviewed chats are disconnected from the account and retained for up to three years. Turning Keep Activity off or using temporary chat does not end all service and safety processing. |
ChatGPT: history can remain after the training switch is off
For a personal ChatGPT account, open Settings, then Data Controls, and inspect “Improve the model for everyone.” OpenAI says that switching it off prevents new conversations from being used to train ChatGPT, while those conversations still appear in chat history. Therefore, use the separate delete controls if your goal is to remove visible history.
Temporary Chat is a narrower mode for a one-off conversation. OpenAI says it does not appear in history, create memories, or improve models, but is retained for 30 days for safety before deletion. That 30-day period means “temporary” does not mean “erased as soon as you close the tab.” Business, Enterprise, Edu, and Healthcare workspace content is not used for training by default, and OpenAI says business products and the API are not trained on inputs or outputs by default unless an organization explicitly opts in. Workspace retention and administrator access still depend on the organization’s configuration.
Claude: deletion and improvement retention are different clocks
Claude Free, Pro, and Max users should inspect the privacy or model-improvement choice. Anthropic says it may use chats and coding sessions to improve models when the user allows that use, otherwise opts in, or when a conversation is flagged for safety review. A safety flag is therefore an exception to a normal preference.
Deleting a Claude conversation removes it from chat history immediately and from backend storage within 30 days. If improvement use is enabled, Anthropic says de-identified data in training pipelines may remain for up to five years. This does not mean every deleted chat is kept for five years; it means eligible, de-identified material already in an improvement pipeline can have a longer retention period than the original conversation record.
Free tools Windows power users keep installed
One-click scans. No signup required.
Rank #2
Microsoft Copilot: separate personal and managed experiences
For personal Copilot, open the profile menu, choose Privacy, and inspect “Training on conversation activity.” Microsoft describes a separate setting for voice conversations. Opting out excludes future conversation activity from training these AI models, but Microsoft says it does not exclude other general product-improvement, advertising, digital-safety, security, or compliance uses.
Microsoft states that personal Copilot’s default conversation-history retention is 18 months. You can delete individual items or all history, but deletion and the training opt-out address different purposes. A work or school Copilot Chat session is different again: enterprise data protection means prompts, triggered Bing queries, and responses are not used to train foundation models by default, while those same records are logged and can be viewed by IT administrators through search and audit tools.
Gemini: Keep Activity, temporary chats, and feedback
Open Gemini Apps Activity and inspect Keep Activity, auto-delete, and the separate setting for audio or Gemini Live recordings. When Keep Activity is on, chats and uploads are saved in activity and Google uses that activity to provide, develop, and improve services, including training generative AI models.
When Keep Activity is off, future chats do not appear in activity and are not used to train Google AI unless you submit feedback. Google still retains those chats for 72 hours to respond and protect users. Temporary chats are not used to train Google AI with human reviewers, but temporary mode does not eliminate all service and safety processing. Google also says a subset of chats may be reviewed by human reviewers; reviewed chats are disconnected from the account and retained for up to three years.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsDoes turning off training delete chats or prevent all access?
No. A training control typically governs a particular future use, not every kind of retention or access. For example, turning off model improvement in ChatGPT leaves new conversations in history, while Gemini with Keep Activity off still retains future chats for 72 hours for service response and protection.
Deletion also has its own timing and exceptions. Anthropic says a deleted Claude conversation disappears from history immediately and backend storage within 30 days, while de-identified data already in training pipelines may be retained for up to five years when improvement use is enabled. Google says chats reviewed by human reviewers may be retained, disconnected from the account, for up to three years. These are provider-specific disclosures, not a universal deletion rule.
When deleting, look for separate controls for chat history, activity, saved memories, and recordings. Read what the provider says about backups, safety retention, and already de-identified data; removing a conversation from the interface is not proof that every related copy or derived data has been erased. If you need an audit trail, note when you changed the setting and when you deleted the material.
Can employees or administrators read my chatbot conversations?
They may be able to, depending on the account and product. Microsoft says work or school Copilot Chat prompts, triggered Bing queries, and responses are logged, and IT administrators can use search and audit tools to view that information—even though those prompts and responses are not used to train foundation models by default.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →This is why “not used for training” should not be read as “private from my employer.” Before using a workplace chatbot, check your organization’s policy and ask what administrators can access, how long logs are kept, and whether your employer has enabled any additional data handling rules. The provider disclosures summarized here do not establish the administrator-access rules for every other chatbot or workspace.
Personal and managed accounts can also look similar in a browser. Confirm the signed-in identity, workspace banner, or account switcher before entering sensitive material. If your employer supplies the account, assume the organization may have governance and audit obligations unless its policy clearly says otherwise.
What should I avoid putting into a chatbot?
Until you have confirmed the provider’s current controls and the account type, do not paste passwords, secrets, regulated personal data, or confidential work material. A training opt-out alone does not settle questions about history, safety review, legal retention, feedback, or administrator access.
- For a one-off sensitive question, use a temporary-chat option where available, but first read what it excludes and how long the provider may retain it.
- Avoid submitting thumbs-up or thumbs-down feedback on a sensitive conversation until you understand how feedback is retained and reviewed. On Gemini, feedback is specifically an exception to the Keep Activity-off rule for future training use.
- Remove identifying details from a prompt when you can. Substitution does not replace checking the provider’s controls, but it limits the consequence of an unexpected retention or review event.
- For work or school use, follow your organization’s approved-data rules and use the account or service it has authorized.
- Do not assume that closing a tab, deleting a browser history entry, or using a browser’s private window changes the chatbot provider’s own records.
What is the safest way to interpret a privacy setting?
Read the setting narrowly and apply it only to the use it names. “Not used for training” concerns model training; it does not necessarily control storage, human review, memory, service operation, safety checks, legal obligations, feedback workflows, or employer access.
There is no single industry-wide rule established by these provider disclosures. The practical answer depends on the live settings, your product and account type, region, and—in a work or school account—administrator policy. Recheck the actual account controls rather than relying on a general statement about a provider’s consumer service.
Rank #4
A useful final check is to write down six answers: Is model improvement on? Is history or activity on? Is memory on? What does temporary mode exclude? Can feedback or safety review create an exception? What is the deletion and access policy? If any answer is unclear, treat the chat as potentially retained and avoid confidential content until you have a clearer answer.
Frequently Asked Questions
Can a chatbot use my conversation after I submit a thumbs-up or thumbs-down?
It may: Google says that when Gemini Keep Activity is off, submitting feedback is an exception to the rule that future chats are not used to train Google AI. Before rating a sensitive conversation, check the provider’s current disclosure on how feedback is handled.
Does deleting a chat remove information the chatbot has already learned?
The documented deletion policies do not promise that deleting a chat reverses every prior use or removes every derived or de-identified record. Anthropic specifically says eligible de-identified data in training pipelines may remain for up to five years when improvement use is enabled.
Recommended Free Tools
Are AI chatbot privacy settings the same in every country?
Not necessarily. Controls and disclosures can vary by region, age, product surface, plan, and administrator policy, so confirm the settings shown for your own signed-in account.
Is a temporary chat the same as an incognito browser window?
No such equivalence is established by the provider disclosures here. Temporary-chat features have provider-specific effects and retention exceptions, so check what the chatbot says the feature does rather than assuming browser incognito behavior applies.
If I turn off ChatGPT model improvement, are my old chats deleted?
No. OpenAI says new conversations remain in chat history after the setting is turned off, although they are not used to train ChatGPT. Delete existing chats separately if you want them removed from visible history.
Best Value
How long does Gemini keep chats when Keep Activity is off?
Google says it retains future chats for 72 hours to respond and protect users. They do not appear in activity and are not used to train Google AI unless you submit feedback, but that does not eliminate all service and safety processing.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Can my employer see a work Copilot conversation even if it is not used for training?
Yes, it may be able to. Microsoft says work or school Copilot Chat prompts, triggered Bing queries, and responses are logged and available to IT administrators through search and audit tools, while not being used to train foundation models by default.
What happens when I delete a Claude conversation?
Anthropic says it is removed from chat history immediately and from backend storage within 30 days. If you allowed model improvement, eligible de-identified data already in training pipelines may be retained for up to five years.
Does turning off Gemini Keep Activity also stop audio recordings?
Not automatically. Google provides a separate setting for audio or Gemini Live recordings, so inspect that control in addition to Keep Activity.
Is a business chatbot always private from the provider?
No universal rule applies. OpenAI says Business, Enterprise, Edu, and Healthcare workspace content is not used for training by default, but retention and organizational access still depend on the product and workspace controls. Check the specific business policy and administrator settings.
The Bottom Line
Check training, history, memory, temporary mode, feedback, deletion, and administrator access as separate questions. A model-training opt-out is useful but narrow: it usually changes future improvement use, not every form of storage or review. Confirm the live settings for your account before sharing anything sensitive.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
