Usually the business, provided the chats happened in a company workspace. On ChatGPT Business, Claude Team or Enterprise, Microsoft 365 Copilot and Gemini in Workspace, the account and its history are the organisation's to manage. Owning it isn't the same as reading it, though, and if the employee used a personal login, the history leaves with them.
How much you can actually get at varies sharply by product. A Claude Team Primary Owner can export every member's chats, including a leaver's. A ChatGPT Business owner can't read or export members' private chats at all; OpenAI keeps a removed member's chats indefinitely, but only so they reappear if that person is re-added. Microsoft keeps Copilot prompts in the user's mailbox, so they follow your email retention rules.
Company workspace or personal login decides most of it
Start with which account the work happened in, because that settles more than any contract clause. Both OpenAI and Anthropic describe business accounts as organisation-managed. OpenAI's notice for managed accounts says the administrator manages the account and the ChatGPT data associated with it, and Anthropic says the Primary Owner of a Team or Enterprise plan manages each member's account and all associated data. The data sits under the company's agreement with the vendor, not the employee's.
A personal account works the other way round. Picture an illustrative landscaping firm whose garden designer spent a year on her own ChatGPT Plus subscription: planting plans, client emails, a pricing template she'd refined over dozens of jobs. When she left, the firm had no route in at all. The template, and the client names and addresses sitting in those chats, stayed in an account the business never controlled. The work may legally be the firm's under her employment contract, but the practical position is that it walked out of the door, along with customer personal data the firm is still responsible for. That is the strongest argument for company AI accounts instead of personal logins, and it applies to a two-person business as much as a large one.
The legal side of ownership (who owns work an employee produces, and what they may take with them) depends on the employment contract and local law, so a short conversation with an HR adviser is worth having if a departure is contentious. The rest of this tutorial deals with the practical side: what each platform lets you keep.
What happens to a leaver's chats, product by product
This table summarises each vendor's current documentation for business plans.
| Product | Who can see a member's chats | Export | When the member is removed |
|---|---|---|---|
| ChatGPT Business | Only the member, plus anyone they share a specific chat with | Not available, for members or owners | Access ends at once; chats, files and canvas documents are kept indefinitely and restored if they're re-added; their projects pass to a workspace owner, but their conversations and files inside them don't become visible |
| ChatGPT Enterprise | Workspace owners through the Compliance API, with the Conversation messages permission | Through the Compliance API | Follows the workspace retention policy, which can be indefinite or a set period such as 90 or 180 days |
| Claude Team | The Primary Owner, through an organisation export | Primary Owner only; includes removed members | Colleagues lose access to the leaver's chats, even ones shared with the organisation; projects shared with everyone stay under the Team tab; private projects become inaccessible |
| Claude Enterprise | As Team | As Team | As Team, plus any custom retention period (minimum 30 days) |
| Microsoft 365 Copilot | Compliance admins, through eDiscovery and Content search | Through eDiscovery | If the account is deleted, messages covered by a retention policy move to an inactive mailbox and stay searchable |
| Gemini in Google Workspace | Vault admins, on editions that include Vault (Business Plus among the business editions) | Vault export | Follows the admin's auto-deletion setting and any Vault retention rules |
The ChatGPT Business row surprises most owners. OpenAI's own pages say plainly that data export isn't available in a Business workspace, and that by default admins and owners can't see all private member chats. So the only way to keep a leaver's useful work is to move it while they are still there, which is exactly what the next example does.
An estimator hands in her notice at a nine-person roofing firm
Consider a nine-person roofing contractor, an illustration rather than a real firm: the owner, an estimator, an office administrator and six roofers. The office team shares three ChatGPT Business Standard seats at $25 a month each. The estimator has used hers for 14 months and has roughly 600 conversations, including the prompts that produce the firm's quote cover emails, insurance-claim summaries and storm-damage updates to customers. She gives two weeks' notice.
The owner can't open her chats or export them, and once she is removed they'll sit in OpenAI's systems, invisible to everyone, unless she is re-added. So they plan the handover inside the notice period:
- Day 1, pick what's worth keeping. Together they list the six jobs she does with AI every week. Of 600 conversations, about 25 matter.
- Days 2 to 5, share and copy. She sends the owner share links for those 25 chats. OpenAI's pages say that when a viewer continues a shared conversation, a new private conversation is created in the viewer's own history, so the owner continues each one briefly and ends up holding his own copies.
- Days 6 to 8, turn chats into prompts. She rewrites the eight most useful prompts as templates in a shared document (the method is in the next section).
- Day 9, rebuild the reference material. The owner creates a project he owns and uploads the current price list, standard terms and warranty wording himself, because files the leaver created won't become visible to him after she goes.
- Last day, remove and reassign. She is removed from the workspace, and the seat goes to the new estimator starting in three weeks.
The whole handover cost about six hours of her time and two of the owner's. The seat arithmetic matters too: removing a member frees the seat but doesn't reduce the bill, and seat reductions only take effect at the start of the next billing cycle. Had the firm not been replacing her, the owner would schedule a reduction from three seats to two, taking the bill from $75 to $50 a month from the next cycle.
Had the firm been on Claude Team instead, the Primary Owner could have exported every chat, the leaver's included, from organisation settings under Data and privacy. The download link arrives by email and expires 24 hours after delivery. That is a real advantage, but an export is a raw archive of conversations, not a usable prompt library, so the handover in steps 1 and 3 would still be the part that saves the next estimator's time.
Turning a leaver's best chats into prompts the team can reuse
A long chat history is a poor handover document. A few tidy, reusable prompts are a good one. The leaver can do most of the work in the AI tool itself, with a prompt like this:
Below is a conversation I used to write quote cover emails for
roof repairs. Turn the final version into a reusable prompt
template for a colleague who has never seen this chat.
- Put anything that changes per job in [square brackets]
- Keep the tone and structure rules I settled on
- List the details the colleague must provide before running it
- Add two lines on mistakes to check for in the output
[paste the conversation]A typical result (illustrative):
PROMPT TEMPLATE: Quote cover email (roof repairs)
Write a cover email for the attached quote.
Customer: [first name]. Property: [street name only].
Work: [one-line summary]. Price: [total inc. tax].
Valid until: [date]. Scaffolding needed: [yes/no].
Rules: under 150 words; open with the job, not a greeting
paragraph; one sentence on what happens next; no discounts
unless [discount] is filled in.
Check before sending: price matches the quote PDF; date is
a working day; no mention of other customers' jobs.Two fixes are usually needed. First, the model often invents a detail from the old conversation, such as a specific tile type that belonged to one job, so read the template for anything that shouldn't be fixed text. Second, the "check before sending" lines tend to be generic; ask the leaver to replace them with the mistakes they actually caught over the past year, because that knowledge is exactly what disappears when someone leaves.
Personal data inside work chats
A leaver's history isn't only business material. It usually contains customers' names, addresses and complaints, and sometimes the employee's own personal questions asked in a work account. An illustrative cleaning company supervisor might have used the company's Claude Team account to draft a letter to her landlord and to ask about sick-pay rules. If the Primary Owner exports the organisation's data after she leaves, those chats are in the file.
Two principles keep this proportionate. Don't trawl a leaver's history out of curiosity: review only what you need for the handover, a customer complaint or a legal request, and write down why. And don't keep everything forever "just in case"; data-protection law such as the GDPR expects personal data to be kept no longer than necessary, which is what retention settings are for. Employees can also ask for copies of personal data held about them, so assume anything in a work chat could one day be read by the person who wrote it. If you're planning to use a leaver's chats in a dispute or disciplinary matter, take advice first.
Policy lines that settle ownership before anyone leaves
Most of the friction above disappears if staff know the rules from their first week. These lines can sit in your AI policy; the full structure is in how to write an AI usage policy for a small business.
AI ACCOUNTS AND CHAT HISTORY (illustrative wording)
1. Company AI accounts are provided for work. Prompts, outputs,
files and chat history in them are company records.
2. Don't use personal AI accounts for company work or for any
customer information.
3. Keep reusable prompts in the shared prompt library, not only
in your own chat history.
4. When you leave, part of your notice period will be spent
handing over prompts and key chats. Don't copy company chats
to personal accounts or devices.
5. We keep AI chat history for [90 days / 12 months]. We may
review it for [security incidents, customer complaints,
legal requests], following our data-protection policy.
6. Personal use of company AI accounts is [allowed within
reason / not allowed]. Administrators may be able to see
personal chats made in work accounts.Have an HR adviser check the wording against your employment contracts, especially line 5. The point of writing it down is that nobody is surprised later: the estimator in the example above knew from day one that her prompts belonged in the shared document, so the handover was a tidy-up rather than a rescue. For the wider leaving process, including shared passwords and connected apps, work through the staff offboarding checklist for AI tools and shared accounts.
When the person leaving holds the owner role
The riskiest departure isn't the heaviest AI user; it's whoever set the accounts up. An illustrative HVAC installer's office manager opened the firm's Claude Team plan with her own work email, which made her the Primary Owner, and she was also the only owner of the ChatGPT Business workspace. Nobody noticed until she left. On Claude, only the current Primary Owner can hand that role to someone else, and an organisation can have just one. On ChatGPT Business, an existing owner has to invite the new account and assign the owner role, and OpenAI notes that creating a fresh workspace doesn't carry over the old one's subscription, members or data. With no owner left in either, the business owner had to ask each vendor's support team to confirm who owned the company, and in the meantime nobody could change billing, remove members or run an export.
The fix takes ten minutes if you do it early:
- Make the business owner an owner on every AI workspace, alongside anyone who manages it day to day.
- On Claude, move the Primary Owner role to the business owner through organisation settings, then Members, then the Role menu; the current Primary Owner confirms by typing the new owner's email address.
- Remember that on Claude an owner can't remove themselves; another owner has to, which is one more reason to have two.
- Put "transfer AI workspace ownership" at the top of the leaving checklist for anyone who holds an owner or admin role.
There is one last-resort option on ChatGPT Business. Because OpenAI restores a removed member's content when the same account is re-added, a firm that controls the leaver's company email address could in principle recover their history later by re-adding that address. Treat it as an exception for a genuine need such as a customer dispute, write down why, and take advice before doing it, because signing in to a former employee's account raises privacy questions of its own.
Settings to change now so the next departure is simple
A few settings, changed once, make every future departure easier. Here is what to do on each platform:
- ChatGPT Business. Keep recurring work in projects owned by the business owner, and accept that there is no export. Handover happens through share links and rebuilt prompts during the notice period, so build that time into your leaving process. If you ever need an owner-level record of every conversation, that's an Enterprise feature through the Compliance API.
- Claude Team or Enterprise. Make sure the Primary Owner is someone who isn't likely to leave, usually the business owner. Encourage staff to share working projects with the organisation rather than keeping them private, because private projects become inaccessible when their creator is removed. On Enterprise, set a custom retention period.
- Microsoft 365 Copilot. Create a Purview retention policy that covers Copilot interactions, so a leaver's prompts are kept for a defined period rather than indefinitely or not at all. Copilot activity is stored in a hidden mailbox folder, so it follows the same leaver process as email.
- Gemini in Workspace. In the Admin console under Generative AI, then Gemini for Workspace, set how long side-panel conversations are kept: 90, 540 or 1,080 days, or no automatic deletion. For the separate Gemini app, Vault retention rules on Business Plus take precedence over admin and user deletion settings.
- Every platform. Keep a shared prompt library outside any one person's chat history. It is the one handover document that survives a change of tool as well as a change of staff.
Anyone who has used AI tools seriously for a year has built up working knowledge in their chat history: phrasing that works with your customers, checks that catch your common errors, formats your clients expect. The business owns the account, but that knowledge only stays if you plan for it before the resignation letter arrives. If you'd like a second pair of eyes on how your accounts are set up, that's part of what my AI implementation consultation covers.
Leavers and AI chat history: follow-up questions
Can a former employee still see their work chats after leaving?
Not if the chats were in a company workspace. Removing someone from ChatGPT Business revokes their access immediately, and a person removed from Claude Team or Enterprise loses access to the organisation's chats and projects. If they used a personal account for work, though, everything stays with them, which is the main reason to insist on company accounts.
Can I read a current employee's AI chats?
It depends on the product and the law. ChatGPT Business gives owners no way to read members' private chats. A Claude Team Primary Owner can export organisation data including chats, and Microsoft and Google offer compliance search tools. Tell staff in writing what may be reviewed and why, and ask an HR or data-protection adviser before reviewing anyone's history.
Should we delete a leaver's AI history once the handover is done?
Keep what your records schedule says you need and let the rest go. Claude Enterprise, Copilot through Purview and Gemini in Workspace all have retention settings that delete automatically. ChatGPT Business keeps a removed member's chats indefinitely, so if your schedule requires deletion, ask OpenAI support how to remove them rather than assuming it happens.
Further reads
- Where Is Your Data Stored When You Use AI Tools? — Where that chat history physically sits, tool by tool.
- ChatGPT Temporary Chat: When Your Staff Should Use It — When staff should use temporary chat instead of saved history.
- How to Share AI Tool Logins Safely With a Password Manager — Why shared logins make ownership and offboarding harder.
- Standard vs Premium AI Seats: Who Needs the Bigger Plan? — Seat choices worth revisiting whenever someone leaves.
- How to Onboard New Hires Onto Your AI Tools and Rules — Set the replacement up with the prompt library from day one.
- Zero Data Retention: What It Means When You Choose an AI Tool — The retention question from the vendor's side of the contract.
- What to Check in an AI Tool's Privacy Policy and Terms — Seven checks for any AI tool's privacy policy and terms, the words to search for, and how ChatGPT, Claude, Gemini and Copilot compare.
- How HR Consultants Use AI for Policies, Letters and Cases — Three workflows for HR consultants, with a client fact sheet, letter before-and-after, a grievance chronology prompt and the lines AI must never cross.
- Does ChatGPT Train on Client Data? Business vs Free Plans — Compare Free and Business data handling, check the account you use, and practise drafting with less client information.
- AI Vendor Lock-In: How to Keep Your Data and Prompts Portable — Where AI lock-in really comes from, which plans can export what in 2026, and a costed exit drill that keeps your prompts and data movable.
- Who Owns the AI Workflows a Consultant Builds for You? — Legal ownership and practical control of AI automations: assignment vs licence, whose accounts they run in, platform transfer rules and sample clauses.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: OpenAI help pages (Data retention when a member is removed from a workspace; Managing data, sharing, and privacy in ChatGPT Business; ChatGPT Business general FAQ; Exporting your ChatGPT history and data; Compliance Platform for Enterprise and Edu); Claude help pages (removing users from Team or Enterprise; exporting organisation data; who owns and manages team data; custom retention for Enterprise); Microsoft Learn (retention for Copilot and AI apps; Data, Privacy, and Security for Microsoft Copilot); Google Workspace Help (Gemini in Workspace conversation history settings; retaining Gemini app messages with Vault; edition comparison). Checked September 2026.