Temporary Chat removes normal history, memory and model-training use, but OpenAI may retain a copy for up to 30 days and business confidentiality rules still apply. It reduces one category of risk — it does not approve the information entered or remove data-processing obligations.
A manager needs help drafting an employee letter. Or a customer complaint response. Or a contract summary, a financial explanation, a supplier dispute or a disciplinary note. They open ChatGPT, start typing — and then notice the Temporary Chat option. They assume: this must be safe for confidential information.
Temporary Chat does offer stronger privacy controls than an ordinary consumer chat. The conversation does not appear in normal history. It does not create or use saved memories. It is not used to improve OpenAI's models. These are genuine and useful protections.
A privacy feature reduces one category of risk. It does not approve the information you put into it.
What Temporary Chat does not do is remove data processing, temporary retention, organisational policy, contractual confidentiality, data-protection obligations or the risk of including unnecessary personal information. Understanding the distinction matters.
Temporary Chat reduces persistence and personalisation. It does not turn a public AI service into a confidential meeting room.
The Quick Answer
Is Temporary Chat Safe for Sensitive Business Discussions?
ChatGPT Temporary Chat is more private than a normal consumer chat because it does not appear in chat history, does not use or create saved memories and is not used to improve OpenAI's models.
However, OpenAI may retain a copy for up to 30 days for safety purposes, and enabled Custom Instructions can still apply.
For business use:
- Remove names and identifying details before submitting
- Never enter passwords, credentials or authentication information
- Follow your organisation's AI policy
- Use an approved business workspace where available
- Do not assume Temporary Chat satisfies legal or regulatory requirements
- Check which connected tools are enabled before starting
Temporary Chat is a privacy control — not permission to disclose confidential information.
What Temporary Chat Is
Temporary Chat creates a conversation with reduced continuity and personalisation. The user selects it before starting a conversation — the precise location of the option varies by device and platform version. Once selected, it produces a session that is handled differently from a standard chat.
It is useful where someone wants a clean conversational context, no visible history entry, no new saved memory, no use of existing saved memories and no use of the conversation for model improvement. Those are meaningful controls for many common situations.
What it is not: incognito browsing, end-to-end encrypted messaging, a secure data room, guaranteed immediate deletion or a replacement for an approved business AI platform.
Temporary describes how the conversation is handled over time — not whether the information is appropriate to submit.
What Temporary Chat Protects
What Temporary Chat Does for Your Privacy
| Protection | What it means |
|---|---|
| No normal history entry | The conversation does not appear in the user's standard chat history list. |
| No new memory created | Information from the conversation is not added to saved memory. |
| No existing memory personalisation | Normal saved memories are not used to personalise the chat. |
| No model improvement | OpenAI currently says Temporary Chats are not used to improve its models. |
| Blanker context | The conversation begins without ordinary previous-chat awareness. |
These controls reduce accidental later discovery in the history list, future personalisation from the discussion, use of the conversation for consumer model improvement and irrelevant influence from previous conversations.
What It Does Not Protect
Temporary Chat does not automatically:
- Anonymise names or remove personal data from the text
- Redact documents or files before processing
- Prevent OpenAI from processing the prompt to generate a response
- Guarantee immediate deletion or override the up-to-30-day safety retention period
- Override enabled Custom Instructions
- Override organisational AI policy or data-protection law
- Override contractual confidentiality obligations
- Prevent the user from taking screenshots or copying the output elsewhere
- Stop connected third-party tools from receiving information
- Prove regulatory compliance for any sector or jurisdiction
- Make passwords or credentials safe to enter
- Make client information appropriate to disclose
- Replace an approved enterprise AI environment
Temporary Chat changes the storage and personalisation behaviour. It does not change the sensitivity of the data.
History, Memory, Training and Retention Explained
These four concepts are often confused. They address different stages of the information lifecycle.
| Concept | Normal chat | Temporary Chat |
|---|---|---|
| Chat history | Conversation appears in the account list | Does not appear in normal history |
| Memory | Can use and create saved memories (if enabled) | Does not use or create saved memories |
| Model training | Eligible conversations may be used for improvement (consumer) | Not used to improve OpenAI's models |
| Custom Instructions | Apply where enabled | May still apply where enabled |
| Data retention | Per standard account and plan terms | OpenAI may retain a copy for up to 30 days for safety purposes |
| Data processing | Required to generate a response | Required to generate a response |
Privacy settings address different stages of the information lifecycle. Turning off one does not automatically turn off the others.
Not saved in your history does not mean never processed or immediately erased.
How to Start a Temporary Chat
The interface changes frequently. The following reflects current official guidance but should be verified before relying on it. Always confirm the Temporary Chat indicator is visible before entering any sensitive information.
Last checked: 2 August 2026. Interface locations change. Verify current steps on the official OpenAI support pages before relying on them.
| Platform | Current approach |
|---|---|
| Web (chatgpt.com) | Click the model name or the pencil/new-chat icon and look for the Temporary option. A visible indicator should confirm the mode is active before you start typing. |
| Windows app | The option is accessible from the new-chat controls. Verify the current menu location in the app before use, as it may differ from the web interface. |
| macOS app | OpenAI currently documents selecting the model name and enabling the Temporary Chat toggle from that menu. Verify current placement before use. |
| Android | Access from the new-chat menu or the options available when starting a conversation. Verify current navigation in the app. |
| iPhone and iPad | Access from the new-chat or menu options. Verify current navigation in the iOS app before use. |
Temporary Chat is available on ChatGPT Free, paid consumer plans, ChatGPT Business and ChatGPT Enterprise. Availability and exact feature scope may vary by plan and region — verify current details on the official OpenAI website. The feature is available in the UK at the time of checking.
Custom Instructions and Safety Context
Custom Instructions are standing instructions a user has supplied to ChatGPT — such as job role, company context, preferred writing style or location preferences. Temporary Chat may still follow these instructions where they are enabled. This means the conversation is not necessarily free from all previously supplied user context.
If Custom Instructions contain information the user is concerned about — for example, a company name, role or specific preferences that could be connected to the topic — it is worth reviewing and temporarily adjusting them before starting a sensitive conversation.
Temporary Chat does not create memory, but enabled Custom Instructions may still shape the conversation.
OpenAI also notes that limited context from prior conversations may be used for safety and security purposes in certain situations. This is distinct from normal personalisation. It does not mean OpenAI routinely reads prior conversations during Temporary Chat. Safety systems may operate separately from the normal memory and history controls, and privacy features do not disable abuse prevention or safety protections.
Uploaded Files and Connected Tools
Users may be tempted to upload complete documents to Temporary Chat — contracts, HR letters, customer lists, financial reports, cyber incident reports or legal correspondence. Before uploading any file, consider:
- Is the full file needed, or only one relevant paragraph or section?
- Can all names be removed before uploading?
- Does the file contain document metadata, tracked changes, comments or hidden worksheets?
- Does it contain passwords, API keys or authentication information?
- Is the account approved for this type of information?
- Is the organisation authorised to use this service for this type of data?
A temporary conversation does not make the document temporary everywhere it may be processed.
Some ChatGPT features can connect to external services including cloud storage, email, calendars, business applications and custom GPT actions. Where a Temporary Chat invokes an external service, information may be handled under that service's separate data-handling terms. Temporary Chat controls OpenAI's conversation handling. It cannot automatically control what an external service retains.
- Review which tools and connectors are enabled before starting
- Avoid invoking external actions unless necessary
- Confirm where information will be sent before approving an action
- Restrict permissions to what the task requires
- Use only approved connectors
- Check the privacy terms of any third-party service involved
What Businesses Should Never Enter
Regardless of whether Temporary Chat is enabled, certain information should not be entered into any public AI service. Exact restrictions depend on organisational policy, sector and the specific approved service in use. As a minimum, do not enter:
- Passwords, one-time codes or authentication tokens
- API keys or private encryption keys
- Recovery codes or banking credentials
- Full payment-card information
- Unredacted identity documents
- Confidential customer databases
- Protected health information without specific approval
- Legally privileged advice without specific approval
- Live cyber-security credentials
- Highly sensitive employee performance or disciplinary information
- Unpublished merger, acquisition or commercial-strategy information
- Secret product designs or technical specifications
- Information prohibited by a contract or non-disclosure agreement
- Safeguarding information
- Information subject to export controls
When Temporary Chat May Be Appropriate
Temporary Chat is useful for tasks where someone wants less persistence and personalisation, provided the information itself has been appropriately reduced. Potential lower-risk uses include:
- Rewriting anonymised text without retaining the conversation in history
- Exploring a sensitive topic using fictional names and placeholder figures
- Asking a general HR-process question without identifying any individual
- Summarising a redacted extract of a document
- Drafting a generic complaint response from an anonymised description
- Discussing a business scenario using label names such as Customer A or Supplier B
- Exploring personal ideas the user does not want added to memory
- Testing prompts without creating a persistent history entry
Reduce the information to the minimum needed for the task.
How to Reduce Information Before Using AI
The following process applies before starting any sensitive AI discussion, with or without Temporary Chat enabled.
- 1Classify. Determine whether the information is public, internal, confidential or restricted — and whether it identifies a person.
- 2Check policy. Confirm whether this AI service is approved for this type of information.
- 3Remove. Delete names, addresses, account numbers and unnecessary identifying details.
- 4Reduce. Supply only the relevant paragraph or figures, not the full document.
- 5Generalise. Replace exact values with ranges or approximate figures where the task permits.
- 6Pseudonymise. Use placeholder labels such as Customer A, Employee B or Supplier C in place of real names.
- 7Check files. Before uploading, remove metadata, tracked changes, comments and hidden worksheets.
- 8Enable Temporary Chat. Confirm the mode is active before entering any information.
- 9Avoid external actions. Do not invoke connected tools or external services unless necessary and approved.
- 10Verify and close. Review the output, save only what is needed and close the session.
The strongest privacy control is often not submitting the information in the first place.
Consumer Accounts Versus Business Workspaces
Individual use of a personal ChatGPT account is subject to consumer terms and is outside organisational control. Businesses should distinguish clearly between account types and define which are approved.
| Account type | Key considerations |
|---|---|
| Personal free or paid account | Consumer terms apply. Staff may use personal accounts without organisational oversight. Approved use should be defined explicitly. |
| ChatGPT Business | Organisational workspace with admin controls. Model training is off by default for conversations. Provides user management and workspace policies. |
| ChatGPT Enterprise | Enhanced controls, identity integration, compliance capabilities and no model training on data. Verify current feature availability with OpenAI. |
| ChatGPT Edu | Designed for educational institutions. Similar elevated controls. Verify current feature set. |
| API-based service | Data handling depends on the API contract and any data-processing addendum. No default model training in most configurations. |
| Unapproved public AI tool | Outside organisational governance. Staff using these services without approval creates shadow AI risk. |
Temporary Chat can improve an individual session. A managed business workspace addresses the wider organisational problem.
Business Policy and Staff Guidance
A practical AI acceptable-use policy should define three information categories and the rules around each.
| Category | Examples | Typical approach |
|---|---|---|
| Permitted | Public information, approved marketing text, anonymised scenarios, generic drafting, non-confidential research | Approved for use on approved services |
| Restricted | Internal operational information, customer correspondence, staff information, financial reports, contracts, security documentation | Approved services only; data minimisation required |
| Prohibited | Credentials, regulated personal data without approval, confidential client databases, legally privileged advice, safeguarding information, commercially critical secrets | Not permitted regardless of feature or mode |
The policy should also define approved services and account types, file-upload rules, connected-tool rules, review responsibilities, incident reporting procedures, deletion and retention expectations, staff training requirements and the consequences of misuse.
Warning Signs
Pause before using Temporary Chat where:
- The information identifies a real customer, employee or other individual
- A contract prohibits the information from being processed by an external service
- Credentials or authentication details appear anywhere in the document
- The file contains hidden data, tracked changes or embedded comments
- The conversation concerns a live cyber-security incident
- Legal privilege applies to the document or discussion
- Regulated personal data or health information is involved
- A third-party connector is enabled
- The user is signed into a personal account rather than an approved business account
- There is no organisational AI policy covering this type of use
- The user assumes Temporary Chat means the conversation is immediately deleted
- The same task could be completed using fully anonymised information
- Nobody can explain the service's current data-processing terms
Where the consequences of disclosure are serious, convenience is not the deciding factor.
Practical Business Implications
| Implication | What it means in practice |
|---|---|
| Temporary Chat is useful | It reduces history, memory and model-improvement concerns for appropriate tasks. |
| It is not a secure data room | Sensitive data still enters an external service and is still processed. |
| Retention still exists | OpenAI may keep a copy for up to 30 days for safety purposes. |
| Custom Instructions may still apply | The session is not completely detached from all account settings. |
| Data minimisation matters | Removing identifying details before entering information is safer than relying only on a mode switch. |
| Connected tools change the risk | External services invoked during a Temporary Chat may have separate retention rules. |
| Business governance is broader | Organisations need approved accounts, policies, training and controls — not just a feature setting. |
The IT Club View
Temporary Chat is a useful feature. It addresses several genuine concerns about normal chat behaviour: the conversation does not appear in history, it does not feed into saved memory, it does not personalise future responses from this session and it is not used for consumer model improvement. For someone asking a sensitive question who does not want it stored permanently in an account, these controls are real and worthwhile.
The risk is that the word 'temporary' invites overconfidence. Users may assume that a temporary conversation means a safe conversation — and begin entering information they should not. OpenAI's retention of the conversation for up to 30 days for safety purposes, the continuation of Custom Instructions and the requirement to process the data all remain in place.
Temporary Chat is safer for a sensitive topic. It does not automatically make sensitive data safe.
IT Club recommends using approved business accounts rather than personal consumer accounts, classifying the information before starting, removing identifying details, providing the minimum necessary context, avoiding credentials completely, reviewing Custom Instructions for anything that should not carry across, checking which connected tools are enabled, following company policy on AI use, using professional advice or controlled systems for high-risk matters and treating the 30-day retention window honestly rather than ignoring it.
Use Temporary Chat as one privacy layer — not as permission to ignore every other layer.
Plain-English Takeaway
ChatGPT Temporary Chat does not appear in normal chat history, does not use or create saved memories and is not used to improve OpenAI's models. OpenAI may still retain a copy for up to 30 days for safety purposes, and Custom Instructions may continue to apply. Use it for reduced persistence and personalisation, but remove identifying information and do not treat it as approval to upload confidential, regulated or security-sensitive business data.
Related Business Questions
What is ChatGPT Temporary Chat?
Temporary Chat is a ChatGPT feature that creates a conversation with reduced persistence and personalisation. It does not appear in normal chat history, does not use or create saved memories and is not used to improve OpenAI's models.
Does Temporary Chat appear in history?
No. Temporary Chats do not appear in the user's standard chat history list.
Does Temporary Chat use memory?
No. Temporary Chat does not use the user's normal saved memories to personalise the conversation.
Does Temporary Chat create memories?
No. Information from a Temporary Chat is not added to the user's saved memory.
Is Temporary Chat used to train AI?
OpenAI currently says Temporary Chats are not used to improve its models.
How long are Temporary Chats retained?
OpenAI may retain a copy of a Temporary Chat for up to 30 days for safety purposes. This is a platform retention period, not a guarantee of immediate deletion.
Are Temporary Chats immediately deleted?
No. OpenAI may retain a copy for up to 30 days for safety purposes. The conversation does not appear in the user's history, but that is not the same as immediate deletion.
Can OpenAI review Temporary Chats?
OpenAI may review Temporary Chats for abuse monitoring during the up-to-30-day safety retention period. This is separate from normal personalisation or memory.
Do Custom Instructions work in Temporary Chat?
Enabled Custom Instructions may still apply during a Temporary Chat. The session is not completely detached from all account settings.
Does Temporary Chat know previous conversations?
Temporary Chat does not access normal saved memories and begins without ordinary previous-chat awareness. OpenAI notes that limited context may be used for safety purposes in certain situations, which is distinct from normal personalisation.
Is Temporary Chat completely private?
No. It provides stronger privacy controls than a normal consumer chat, but information is still transmitted to and processed by the service. OpenAI may retain a copy for up to 30 days for safety purposes.
Is Temporary Chat anonymous?
No. Temporary Chat does not anonymise the content of what you enter. The user remains authenticated. Names, details and identifying information in the text are visible to the service.
Is Temporary Chat encrypted?
Communications with ChatGPT use standard HTTPS encryption in transit. Temporary Chat does not add end-to-end encryption beyond what applies to all ChatGPT communications. It is not an encrypted messaging service.
Can Temporary Chat be used for confidential information?
Temporary Chat reduces persistence and personalisation, which is useful. But it does not override organisational policy, contractual confidentiality, data-protection law or the need to minimise identifying information. Whether any specific confidential information is appropriate depends on policy, sector and the approved service in use.
Can I upload files in Temporary Chat?
File uploads may be available in Temporary Chat. Any file uploaded is still transmitted to and processed by the service. Files should be stripped of unnecessary personal information, metadata, comments and credentials before uploading.
Are uploaded files deleted after 30 days?
The up-to-30-day safety retention period refers to the conversation. Retention of uploaded files may be governed by separate terms. Check the current OpenAI privacy policy and data-processing documentation for precise wording.
Can Temporary Chat use web search?
Web search may be available during Temporary Chat depending on your plan and settings. Any search query is processed by the service and may be sent to a search provider. Verify current behaviour on the official OpenAI support pages.
Can Temporary Chat use connected apps?
Connected tools and apps may be available during Temporary Chat. Where a connected service is invoked, information may be handled under that service's separate data-processing terms.
Can GPT actions receive Temporary Chat data?
If a GPT with actions is used during a Temporary Chat, those actions may transmit information to third-party services. Temporary Chat controls OpenAI's handling — it cannot control what an external service retains.
Can I share a Temporary Chat?
Shared links for Temporary Chats are not available in the same way as standard chats. Verify current behaviour on the official OpenAI support pages.
Can I recover a Temporary Chat?
Temporary Chats do not appear in normal history and are not designed to be recovered by the user after the session ends. Do not use Temporary Chat for work you need to retrieve later.
Can Temporary Chats be exported?
Standard export features apply to history. Temporary Chats are not included in normal history exports. Users can still manually copy or screenshot the content during the session.
Is Temporary Chat available in the UK?
Temporary Chat is available in the UK at the time of checking. Verify current availability on the official OpenAI website.
Is Temporary Chat available on free accounts?
Temporary Chat is currently available on ChatGPT Free accounts as well as paid consumer plans, Business and Enterprise. Feature availability may vary. Verify current plan details.
Is Temporary Chat available on mobile?
Temporary Chat is available on the Android and iOS apps. The interface location may differ from the web version. Verify current navigation in the app.
How do I start Temporary Chat?
On the web, look for the Temporary option near the new-chat or model-name controls. On macOS, OpenAI currently documents selecting the model name and enabling the Temporary Chat toggle. On mobile, it is accessible from the new-chat or menu options. Interfaces change — always confirm the Temporary indicator is visible before entering information.
Is Temporary Chat suitable for HR matters?
Temporary Chat reduces persistence. It does not remove data-protection obligations relating to employee information. HR-related AI use should follow organisational policy, and identifying information should be removed before any AI discussion.
Is Temporary Chat suitable for legal documents?
Legal professional privilege attaches to confidential communications between lawyers and clients. Temporary Chat does not preserve privilege. Do not enter legally privileged content without explicit approval from legal advisers.
Is Temporary Chat suitable for financial information?
Temporary Chat reduces persistence but does not remove the transmission and processing of the information. Sensitive financial information should be minimised before use, and approved business accounts should be used where financial data processing is involved.
Should passwords ever be entered into AI?
No. Passwords, authentication tokens, one-time codes, API keys and encryption keys should never be entered into any AI service regardless of mode or setting.
What is the difference between memory and history?
History is the list of past conversations visible in the account. Memory is information ChatGPT may save from conversations to personalise future chats. Temporary Chat removes both: it does not appear in history and does not use or create memories.
What is the difference between training and retention?
Training refers to using conversation data to improve future AI models. Retention refers to how long the service keeps a copy of the conversation. Temporary Chat opts out of model training but may still involve retention for safety purposes for up to 30 days.
Is ChatGPT Business safer than a personal account?
ChatGPT Business provides organisational controls, user management, workspace policies and model-training protections that are not available on personal consumer accounts. For business use, a managed workspace is more appropriate than relying on Temporary Chat in a personal account.
How should businesses anonymise information?
Effective anonymisation typically involves removing all direct identifiers (names, addresses, account numbers), replacing them with placeholder labels, removing indirect identifiers that could re-identify someone in context, removing metadata and checking that the remaining information cannot reasonably be traced back to an individual or organisation.
What should an AI acceptable-use policy include?
An AI acceptable-use policy should define approved services, approved account types, permitted information categories, restricted information, prohibited information, file-upload rules, connected-tool rules, data-minimisation expectations, incident reporting procedures, training requirements and the consequences of misuse.
Administrator Technical Note
IT administrators managing ChatGPT use across an organisation should understand the technical distinctions between account types, features and organisational controls.
Account types and workspace controls
Personal free and paid accounts are governed by consumer terms and are outside organisational control. ChatGPT Business and Enterprise workspaces provide admin consoles, user management, workspace policies and model-training protections. Enterprise workspaces may also support SSO (single sign-on) and SCIM for identity management. Administrators should ensure staff use only approved account types and that shadow AI — use of personal accounts for work tasks — is addressed in policy and training.
Temporary Chat, history, memory and training controls
Temporary Chat is a user-level feature. It does not produce workspace-level audit logs or admin visibility. History, memory and model-training controls may also be available at workspace level in Business and Enterprise plans — these operate separately from and in addition to the individual Temporary Chat feature. Administrators should not assume that Temporary Chat adoption by staff replaces workspace-level governance.
Retention, deletion and abuse monitoring
OpenAI may retain Temporary Chats for up to 30 days for safety purposes, including abuse monitoring. The precise retention mechanism and deletion process are governed by OpenAI's current data-processing and privacy documentation. Administrators conducting a data-protection impact assessment (DPIA) or supplier assessment should refer to the current OpenAI data-processing addendum for the relevant plan, not to general consumer documentation.
Custom Instructions, safety context and connected tools
Custom Instructions persist across Temporary Chats where enabled. Safety-related context may also be used in certain circumstances. Connected tools — including GPTs with actions, web search and external app connections — can transmit information to third-party services during Temporary Chats. Each connected service has its own data-handling terms. Administrators should audit which connectors are enabled in the workspace and restrict or approve them accordingly.
File uploads and document metadata
Uploaded files may contain document metadata (author name, organisation, tracked changes, comments, revision history) that is not visible in the rendered text. This metadata is transmitted with the file. Staff should be trained to strip metadata from documents before uploading. Microsoft Office and PDF files can retain significant metadata. Tools for metadata removal should be recommended in policy.
Browser controls, mobile application management and data-loss prevention
Browser-based access to ChatGPT can be controlled via browser policy or content filtering. Mobile access can be managed through mobile application management (MAM) or mobile device management (MDM) policies. Data-loss prevention (DLP) tools may be able to detect and block uploads of certain document types or content patterns to AI services. These technical controls supplement policy but should not be the sole reliance.
Information classification, redaction and pseudonymisation
An information classification scheme — public, internal, confidential, restricted — should drive AI-use decisions. Redaction tools (removing specific text) and pseudonymisation (replacing identifiers with labels) reduce the risk of inadvertent disclosure. Staff guidance should include practical examples of how to reduce information before using AI tools.
Audit logging and incident response
Business and Enterprise plans may provide audit logging for workspace activity. Logs should be reviewed periodically and included in incident-response procedures. If a staff member inadvertently discloses personal data through an AI service, data protection law may require assessment of whether a reportable incident has occurred. AI-related incidents should be included in the standard incident-response process.
Staff training
Training should cover: what Temporary Chat does and does not protect; the data-minimisation process; the prohibited information list; the difference between personal and business accounts; how to report an accidental disclosure; and where to find the current approved AI service list.
Operational Heartbeat
AI privacy needs an Operational Heartbeat: approved tools, account types, retention, memory, connected services, information classification and staff behaviour should be reviewed rather than assumed to remain safe.
A recurring review of AI privacy arrangements should check:
- Approved AI services and whether the list remains current
- Approved account types and whether staff are using only those accounts
- Model-training settings on each approved service
- Memory and Custom Instructions settings in use across the organisation
- Current retention terms for each approved service
- Connected tools and apps — what is enabled and what should not be
- File-upload permissions and staff awareness of metadata risks
- Staff guidance — whether it reflects current platform behaviour
- Information classification scheme — whether it addresses AI use
- Shadow AI — whether personal accounts are being used for work tasks
- Incident reporting — whether any AI-related disclosures have been reported
- Previous corrective actions — whether they remain in place
AI privacy arrangements change as providers update retention terms, memory features change, Custom Instructions evolve, connectors are added, staff create personal accounts, business plans change, app interfaces change, new data types are uploaded, regulations develop, employees leave and approved-use rules become outdated.
Set a review date. Do not assume that what was checked six months ago still accurately describes the current service behaviour or organisational risk.
Plain-English Takeaway
ChatGPT Temporary Chat does not appear in normal chat history, does not use or create saved memories and is not used to improve OpenAI's models. OpenAI may still retain a copy for up to 30 days for safety purposes, and Custom Instructions may continue to apply. Use it for reduced persistence and personalisation, but remove identifying information and do not treat it as approval to upload confidential, regulated or security-sensitive business data.
Need the practical steps?
A short, instruction-led version of this topic is available in the Knowledge Centre.
View the Knowledge Centre GuideRelated Articles
Is Your Business Ready for the Vulnerability Patch Wave?
AI-assisted security tools can search large codebases and identify possible software vulnerabilities much faster than traditional manual research alone. This may create a Vulnerability Patch Wave — a sustained increase in security advisories, emergency fixes and updates that organisations must assess, test and deploy faster than before. This article explains what the wave is, why discovery is accelerating, why fixing remains slower and what businesses should do.
Read articleCan Someone Pretend to Email Your Customers?
A customer receives an email that looks exactly like it came from your business. It asks them to pay an invoice, click a link or reset a password. It did not come from you. This is email spoofing — and DMARC is one of the best tools available to stop it.
Read articleDeleted a OneDrive File? Where You Need to Look for It Now
Microsoft changed how the OneDrive sync client handles cloud file deletions. A file deleted online may disappear from your computer without an additional copy appearing in the Windows Recycle Bin.
Read article