Free tools Windows power users keep installed
One-click scans. No signup required.
For genuinely sensitive work, use only an organization-approved AI workspace whose contract, retention rules, administrator controls, and data-use terms have been checked for your exact account. If you cannot verify those conditions, remove names, identifiers, confidential passages, credentials, and regulated data—or do not submit the material. A chatbot’s brand name is not enough: a personal account and an employer-managed account from the same provider can have different storage, training, review, and administrator-access rules.
This guide shows how to evaluate those differences before you upload a document or type a confidential prompt. It focuses on data handling, not model quality or generic security accessories.
As an Amazon Associate I earn from qualifying purchases.
Start with the account, not the chatbot brand
The first question is not “Which chatbot is smartest?” It is “Which legal and technical environment will receive this data?” Consumer services are normally governed by individual privacy settings and consumer terms. A managed business, education, healthcare, or API deployment may instead be governed by an organization’s contract, data-processing agreement, identity system, retention policy, and audit configuration.
OpenAI explicitly separates personal services from Business, Enterprise, Edu, Healthcare, and API offerings. Microsoft ties Enterprise Data Protection (EDP) to eligible work or school sign-ins rather than to the word “Copilot” alone. Anthropic’s published consumer rules likewise should not be applied automatically to Claude for Work or its API. Check the exact product edition, account owner, region, and subscription before treating any privacy statement as applicable.
#1 Best Overall
If an employer, school, or client provides the workspace, ask the administrator for the product name, subscription, governing contract, retention period, deletion process, logging scope, connected services, and approved-data policy. If you are deciding for yourself, read the current consumer privacy controls and set them before entering sensitive text.
What to evaluate before sending confidential material
1. Model-improvement use
Determine whether prompts, uploaded files, generated responses, feedback, or support records can be used to improve models. “Not used for training” answers only one question. It does not establish that data is not stored, reviewed for safety, visible to administrators, or transferred through a connector.
2. Ordinary retention and deletion
Find out how long normal chats and uploaded files remain, whether deleting a conversation also removes attachments and derived records, and whether backups or legal holds extend the period. Look for different rules for deleted accounts, abuse investigations, support tickets, and feedback.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →3. Safety review and human access
Providers may use automated systems or trained reviewers for abuse detection, safety enforcement, quality work, or troubleshooting. Identify the triggers, the material that can be reviewed, and the stated retention period. A temporary or incognito mode can limit model-improvement use without making a conversation instantly disappear.
4. Memory and personalization
Memory features can carry facts from one conversation into later chats. Check whether memory is enabled, what is saved, how to inspect or delete it, and whether workspace administrators can set a common policy. Turning off model improvement does not necessarily turn off memory.
5. Administrator visibility and audit
In a managed workspace, administrators may be able to search, export, audit, or retain prompts and responses. Ask whether logs include uploaded files, tool calls, web searches, timestamps, user identity, and generated output. Separate “the provider does not train on this data” from “my employer cannot see it”; those are different statements.
6. Identity, permissions, and connectors
Business systems may inherit your organization’s identity, file permissions, sensitivity labels, retention labels, and access revocation. Confirm what happens when a user changes roles or leaves. Review every connector, plug-in, browsing tool, cloud drive, code repository, and external search service because those systems can receive prompt context under their own terms.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
7. Residency and processing region
Some plans offer storage or processing in selected countries or regions; eligibility and supported endpoints vary. “Hosted in Europe” may refer to data at rest but not every support, abuse, or inference operation. Verify the precise region promises in the contract and product documentation.
8. Contractual role and configuration responsibility
Determine whether the provider acts as a processor for your organization, which data-processing agreement applies, and which settings your administrator must configure. A contractual feature is not automatically active: retention, access groups, audit, connectors, and regional controls can require deployment decisions.
Consumer and organizational experiences are not interchangeable
| Provider and experience | Model-improvement treatment | Retention, review, or visibility described in official material | Controls and qualifications |
|---|---|---|---|
| OpenAI consumer services | Submitted content may be used for model improvement depending on settings. | Temporary Chats are not used to improve models while temporary; OpenAI may retain a copy for up to 30 days for safety. | Model improvement: Settings > Data controls. Memory: Settings > Personalization. Deleting or saving a Temporary Chat changes how it is handled. |
| OpenAI Business, Enterprise, Edu, Healthcare, and API | By default, customer content is not used to improve models. | Exact retention and administrator access depend on the product, contract, workspace policy, and endpoint. | Organization data ownership, role/access management, SSO, and other controls are described for business offerings. Residency is available only to eligible customers and products. |
| Anthropic consumer Claude | Chats and coding sessions may be used when the user chooses model improvement, when a conversation is flagged for safety review, or when the user explicitly opts in. | Incognito chats are not used to improve Claude even when Model Improvement is enabled. A feedback-linked conversation may be stored in a secured backend for up to five years. | The published article is for consumer products. Do not transfer its rules to Claude for Work or the API. |
| Microsoft Copilot with work or school sign-in | Prompts, responses, and Microsoft Graph data are excluded from foundation-model training. | Prompts, Bing queries triggered by prompts, and responses are logged and may be viewable by IT administrators through search or audit tools. | EDP, identity, permissions, labels, retention, and auditing vary with the underlying subscription and tenant configuration. |
| Microsoft personal Copilot | Consumer training and review practices vary by user category and location. | Personal privacy controls cover conversation history retention, personalization or memory, training use, and ads; some consumer conversations can receive automated or human review. | Microsoft notes that a newer app version became available on August 18, 2026; follow the current linked controls rather than assuming older-app text applies. |
For any cell that matters to your work, verify the exact plan and region immediately before use. Provider policies can change, and a business label does not establish a particular retention period or administrator configuration.
OpenAI: settings for personal use and controls for managed workspaces
Personal ChatGPT services
OpenAI’s consumer guidance describes controls for model improvement, memory, Temporary Chat, account deletion, and privacy requests. To change model-improvement use, open Settings > Data controls. To manage memory, open Settings > Personalization. Review both areas: disabling one does not necessarily disable the other.
Temporary Chat stays out of chat history, does not create or update memories, and is not used to improve models while it is temporary. OpenAI states: “OpenAI may retain a copy for up to 30 days for safety purposes.” If you later save a Temporary Chat, it becomes a regular chat and follows the account’s ordinary settings. Therefore, use Temporary Chat as a bounded reduction in history and improvement use—not as a promise of immediate deletion or zero human access.
For consumer and individual services, submitted content can be used for model improvement when the applicable settings allow it. Before sharing a sensitive file, remove identifiers and inspect whether memory, feedback, shared links, custom instructions, or connected tools could preserve additional context.
Source: OpenAI consumer data controls and handling.
Business, Enterprise, Edu, Healthcare, and API offerings
OpenAI says content from ChatGPT Business, Enterprise, Edu, Healthcare workspaces, and the API Platform is not used to improve models by default. Its business information also describes customer ownership of organization data, security features, role and access management, and SSO. Those statements do not replace a review of your contract, workspace policy, retention configuration, connected applications, or API endpoint.
Eligible Enterprise, Edu, Healthcare, and API customers can have data-residency choices at rest in the United States, Europe, the United Kingdom, Japan, Canada, South Korea, Singapore, Australia, India, and the United Arab Emirates. Availability depends on eligibility and product. OpenAI also describes in-region GPU inference for eligible customers in the United States or Europe and processing-region selection on supported API endpoints.
OpenAI states that a business-associate agreement is offered to ChatGPT for Healthcare and API healthcare customers to support HIPAA requirements. That offer does not mean every OpenAI product is HIPAA compliant, and using a product alone does not make an organization compliant. Confirm the applicable agreement, enabled features, and internal controls with your legal or security team.
Sources: OpenAI business data protections and residency and OpenAI consumer handling details.
Anthropic: consumer Incognito is narrower than deletion
Anthropic’s Privacy Center article dated March 16, 2026 covers consumer products. It says Claude consumer chats and coding sessions may be used to improve models when the user selects that option, when a conversation is flagged for safety review, or when the user otherwise opts in.
Incognito chats are not used to improve Claude even if Model Improvement is enabled. That is a model-improvement exclusion, not a guarantee that the chat is never stored or reviewed. If you submit feedback, Anthropic says the entire related conversation may be stored in a secured backend for up to five years and may be used for analysis and training as allowed by law.
Because this policy is expressly about consumer products, obtain the separate commercial terms for Claude for Work or the API before treating either as suitable for confidential material. Read the current account controls and determine whether feedback submission could create a longer-lived record.
Source: Anthropic Privacy Center: consumer model-improvement use.
Microsoft Copilot: work/school EDP still permits administrator access
Organization-managed Copilot
Microsoft’s Enterprise Data Protection documentation says organizational Copilot and Copilot Chat use is covered by Microsoft’s Data Protection Addendum and Product Terms, with Microsoft acting as processor. It describes encryption, tenant isolation, privacy commitments, and inheritance of identity, permissions, sensitivity labels, retention policies, auditing, and administrator settings. The exact controls vary by subscription.
Prompts, responses, and data accessed through Microsoft Graph are not used to train foundation models under the organizational EDP description. However, Microsoft Support says Copilot Chat with a work or school account logs prompts, Bing queries triggered by prompts, and responses; IT administrators may view them through Microsoft search and audit tools. Bing operates separately and has different data handling, so a web result or triggered query must be evaluated as another data path.
That means EDP can protect organizational data from foundation-model training while still allowing authorized administrators to investigate or retain activity. Ask your administrator who can access logs, how long they persist, whether eDiscovery or legal holds apply, and which connectors are enabled.
Sources: Microsoft Enterprise Data Protection and Microsoft work/school Copilot Chat data protection.
Personal Copilot
Personal Copilot has separate controls for personalization and memory, conversation use for training, advertising, and conversation-history retention. Microsoft’s privacy-controls page notes that a newer app version became available on August 18, 2026, and that some text applies only to the older app. Use the current controls linked from that page rather than relying on screenshots or legacy instructions.
Recommended Free Tools
Microsoft’s consumer FAQ says some conversations may undergo automated and human review and that training practices and exceptions vary by user category and location. Its direct warning is: “You shouldn’t provide any confidential or sensitive personal data that you would not want Microsoft to use for purposes explained in this FAQ and in the Microsoft Privacy Statement.” Treat a personal account as unsuitable for protected work unless your organization has explicitly approved it and the applicable terms support that use.
Sources: Microsoft Copilot privacy controls and Microsoft Copilot privacy FAQ.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.A practical decision procedure
- Classify the material. Mark whether it contains personal data, health or financial records, legal advice, trade secrets, source code, credentials, customer content, export-controlled information, or information covered by a contract.
- Identify the account. Record the provider, exact product and plan, personal versus work/school sign-in, tenant or organization owner, region, and whether an API or third-party application is involved.
- Read the applicable terms. Confirm model-improvement use, safety-review exceptions, ordinary retention, deletion, feedback handling, memory, support access, and subprocessors. Save the policy version or contract reference used for approval.
- Map every data path. List uploaded files, chat history, memory, connectors, browsing, triggered search queries, telemetry, exports, shared links, and administrator or eDiscovery access.
- Check configuration. Verify model-improvement and memory settings, temporary or incognito mode, retention labels, access groups, logging, regional controls, and connector permissions. Ask who can change each setting.
- Minimize the prompt. Replace names with roles, remove account numbers and secrets, summarize instead of uploading a full document, and provide only the passage needed for the task. Never paste passwords, private keys, authentication codes, or unredacted customer databases.
- Run a low-risk test. With nonconfidential text, confirm which history, memory, audit, export, and deletion behaviors actually occur in your account. A test cannot prove a policy promise, but it can reveal a mistaken plan or enabled connector.
- Obtain approval where required. For regulated or contractual data, have legal, privacy, security, or the data owner approve the precise deployment. Record the permitted data classes and prohibited uses.
- Recheck before material changes. Repeat the review after a plan change, app migration, new connector, regional move, policy update, or change in employer ownership.
How to choose by situation
Private personal planning
Use the provider’s current consumer controls, disable model improvement if available, review memory, and prefer a temporary or incognito mode when you do not need continuity. Do not enter information that would cause harm if retained, reviewed, or exposed through an account compromise. Consumer controls reduce risk; they do not create an employer-grade confidentiality obligation.
Employer documents and internal analysis
Use the employer’s approved managed workspace, not a personal account. Confirm that the plan supplies the required contractual protections and that administrators have configured retention, permissions, connectors, and audit access. Assume activity may be logged in a work environment even when prompts are excluded from model training.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsClient, legal, health, or financial records
Require written approval for the exact product and data class. Verify the processing role, residency, deletion, subprocessors, incident terms, access logs, and any required agreement such as a healthcare business-associate arrangement. Redaction and data minimization remain necessary even when a contract prohibits model-improvement use.
Source code and credentials
Use an organization-approved coding environment with repository permissions and secret scanning. Do not paste private keys, tokens, production connection strings, or complete proprietary repositories into a consumer chat. Check whether coding sessions, feedback, extensions, or external tools create additional copies.
Common mistakes and how to correct them
- “It says no training, so it is private.” Check storage, safety review, administrator visibility, memory, connectors, and residency separately.
- “Temporary means deleted.” OpenAI’s Temporary Chat can be retained for up to 30 days for safety; other products have different rules.
- “Incognito means no record.” Anthropic’s Incognito exclusion concerns model improvement. Feedback can create a conversation record retained for up to five years under its consumer policy.
- “Business is a universal privacy tier.” Plan names do not establish the same contract, region, retention, or admin settings across providers.
- “My employer cannot see work chats.” Microsoft documents that organizational Copilot prompts, triggered Bing queries, and responses may be viewable by IT administrators.
- “The connector is inside the chatbot, so its rules do not matter.” A drive, search engine, repository, or plug-in can receive context under separate permissions and retention terms.
- “A privacy screen or security app solves the problem.” Device protections can reduce shoulder-surfing or malware risk, but they do not determine how a provider stores, reviews, trains on, or exposes prompts.
- “A compliance badge makes my workflow compliant.” Compliance depends on the exact service, agreement, configuration, users, data, and operating procedures.
A short approval checklist
- Exact provider, product edition, plan, account type, tenant, and region recorded.
- Model-improvement use and all exceptions documented.
- Normal retention, deletion, backup, legal-hold, and feedback rules understood.
- Memory and personalization reviewed and configured.
- Safety-review and human-access conditions identified.
- Administrator search, audit, export, and eDiscovery visibility confirmed.
- Connectors, web search, Bing or other external services, and subprocessors mapped.
- Identity, permission inheritance, sensitivity labels, and access revocation tested.
- Residency and processing-region commitments match the requirement.
- Redaction, minimization, incident response, and approval owner documented.
Bottom line for a confidential prompt
Choose the approved organizational workspace whose exact contract and configuration answer every lifecycle question: who receives the data, whether it can improve models, how long it remains, who can review it, whether administrators can audit it, which tools see it, and where it is processed. If any answer is unknown, treat the material as too sensitive for that account and redact it or keep it out of the chatbot.
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.
