Can executive assistants use ChatGPT? What an AI reaches through delegate access
The Morse teamPublished 10 min read
Executive assistants can use ChatGPT, Claude, or Gemini for their own drafting, but as of November 2026 an executive's mail, calendar, board papers, and deal material belong only in the AI tools the employer has approved. Delegate access raises the stakes for an assistant. Microsoft says a delegate can use the same mailbox permissions through Copilot, so the AI on your account can see what your executive lets you see.
This isn't legal advice. On November 18, 2026, we read Google's Gmail delegation help and its Workspace page on what Gemini can reach, Microsoft's delegate, calendar sharing, and Copilot pages, Regulation FD and Rule 10b5-2 in the eCFR, the SEC's amended complaint in SEC v. Meadow, the consumer and business data terms of OpenAI, Anthropic, and Google, and OpenAI's own prompt pack for government executive assistants. Two companies earn credit here. Google keeps Gemini out of delegated inboxes altogether, and OpenAI's prompt pack tells assistants to "Follow your organization's rules for approved products, workspaces, tools, and data."
Is it safe for an executive assistant to use ChatGPT?
For your own writing, yes. For your executive's material, only on a plan your employer runs. OpenAI's pricing page marks Free, Go, Plus, and Pro with "Content is used to train our models" and "Opt-out available," and its page on consumer data says authorized staff and "trusted service providers" may read user content for abuse, support, legal matters, or model improvement, then asks: "Please do not enter sensitive information that you would not want reviewed or used." A temporary chat stays out of your history and isn't used to improve OpenAI's models while it stays temporary, but "OpenAI may retain a copy for up to 30 days for safety purposes." Anthropic says its business terms don't cover "Claude.ai or Claude Pro use for individuals or entities," and its consumer terms allow training "unless you opt out" (Anthropic). Google says human reviewers read some Gemini chats, keeps reviewed ones "for up to three years," and asks you not to enter "confidential information that you wouldn't want a reviewer to see" (Google).
The work plans flip the training default. OpenAI says of ChatGPT Business and Enterprise, "We do not train our models on your data by default" (OpenAI), and Anthropic's commercial terms say "Anthropic may not train models on Customer Content from Services" (Anthropic). They also give your employer a view. OpenAI says Business admins "can view, access, export, and delete end user conversations" (OpenAI), which we cover in is ChatGPT Business private.
OpenAI's own prompt pack for government executive assistants, published August 11, 2026, draws the same line. It says not to enter "classified, controlled, privileged, procurement-sensitive, personal, or otherwise restricted information" unless "your organization has explicitly authorized that use," and its daily brief prompt starts from an "approved calendar export."
Can an AI read my boss's email if I have delegate access?
In Gmail, Google's own AI can't. In Outlook, Microsoft's can. Google's delegation help says a delegate "can read, send, and delete emails in your account for you," and that delegation grants "permission to manage your email only," so "features that integrate with non-email services or sensitive Google Account settings aren't available to delegates." Gemini in Gmail is on that list, along with Smart Compose and Smart Reply. Google's Workspace page on what controls Gemini's access says it again: "If someone has given you delegated access to their inbox, Gemini doesn't have access to those messages."
Microsoft went the other way. Its delegate guide says "Delegates can also use those same item permissions using Copilot (such as to ask for a summary of a message they have access to) if Copilot is enabled for the Delegate's account." Copilot Chat in Outlook can "Summarize and Draft work for all shared and delegated mailboxes" and "Summarize items for all delegate calendars," and you can ask about the mailbox "as if the mailbox were your own" (Microsoft). Copilot can't yet send mail or schedule meetings in a delegated mailbox, which Microsoft lists as "Coming soon."
Microsoft's limit is your own permissions. Copilot "does not elevate access or expose content beyond what the user is already permitted to view," and "Conversation history is saved to your primary account." Its Copilot privacy page says Copilot "only surfaces organizational data to which individual users have at least view permissions," and that prompts and data reached through Microsoft Graph "aren't used to train foundation LLMs." That's a work tool your company licensed, under its contract. Outside apps work through the account you sign in with: OpenAI says ChatGPT's apps respect "your organization's existing permissions" and each user has to authenticate first (OpenAI). What a personal connector holds on your mail is in our guide to connecting Gmail to AI.
What delegate access hands an AI, Gmail vs Outlook
Each cell comes from Google's or Microsoft's own pages, read November 18, 2026.
| Gmail (Google Workspace) | Outlook (Microsoft 365) | |
|---|---|---|
| What a delegate can do | Read, send, and delete mail; can't chat or change the password; up to 1,000 delegates on a work account (Google) | Reviewer reads items; Author also creates items; Editor also changes and deletes the executive's items (Microsoft) |
| The company's AI inside the delegated mailbox | Gemini in Gmail isn't available to delegates (Google) | Copilot Chat summarizes and drafts in delegated mailboxes and summarizes delegate calendars (Microsoft) |
| The company's AI on your own account | Gemini reaches only your own inbox and the Calendar events you can view (Google) | Copilot surfaces data you have "at least view permissions" to (Microsoft) |
| Private items | Not covered on the delegation page | "Let delegate view private events" shows a delegate the details others see only as a time slot (Microsoft) |
| Whose name is on sent mail | The delegate's email address appears (Google) | Send on Behalf shows both names; Send As, set only by an admin, shows the executive's alone (Microsoft, Microsoft) |
| Where the AI chat is saved | No Gemini chat in a delegated inbox | Your primary account; others using the same shared mailbox don't see it (Microsoft) |
Does marking a meeting private hide it from a delegate's AI?
Not if you were given the private items. Microsoft says calendar items marked Private show most people "only the time," not the subject or location, and adds, "The exception to this is if you select the Let delegate view private events check box." In classic Outlook, Delegate can see my private items "affects all Exchange folders," mail, contacts, and tasks included, and "There is no way to grant access to private items in only specified folders." Microsoft adds: "You shouldn't rely on the Private feature to prevent other people from accessing the details of your appointments, contacts, or tasks" (Microsoft).
Put that next to Copilot working within what you're "already permitted to view," and an executive's private doctor's appointment or a confidential deal meeting you can see is within reach of the Copilot on your account. The executive sets that box, so agree with them which items you should see.
What does the SEC say happened when someone got into an executive assistant's laptop?
The SEC alleges that deal information leaked from one through her calendar. In an amended complaint filed September 29, 2026, the SEC says an executive assistant at a New York investment bank supported several bankers who "shared their Microsoft Outlook calendars with her," so "anyone with access to the Executive Assistant's Outlook account would also have access to the calendars and calendar items of these bankers." Deal teams sent her scheduling request forms with "the names of the companies participating in, and the material terms relating to, the proposed deals," which she attached to the meeting invitations (SEC complaint).
The complaint says she worked from the apartment she shared with her partner, Steven Teixeira, and logged in to the bank's systems on her personal laptop. It alleges he gave her a device that moved her mouse so the bank's portal wouldn't lock, read the meeting attachments while she was away, and passed what he found to people who traded on it. The SEC alleges profits of more than $730,000 for one of them, Jordan Meadow (SEC). These are the SEC's allegations, and the assistant isn't a defendant.
No AI was involved. In the SEC's telling, the assistant's account held the shared calendars of the bankers she supported, and the person who got into it saw them too. By Microsoft's own description, Copilot on an account like hers works within that same reach.
Is pasting board or deal information into ChatGPT a securities problem?
The rules' text doesn't name an AI company, but the information is exactly what they protect. Regulation FD applies when an issuer, "or any person acting on its behalf," discloses material nonpublic information to brokers, investment advisers, investment companies, or holders likely to trade on it, and it exempts a person "who expressly agrees to maintain the disclosed information in confidence" (17 CFR 243.100). Rule 10b5-2 says a duty of trust or confidence exists "Whenever a person agrees to maintain information in confidence" (17 CFR 240.10b5-2).
If you've signed a confidentiality agreement with your employer or with a deal counterparty, a board deck or a target's name is information you agreed to keep. On a consumer plan, a copy then sits with an AI company whose staff and reviewers may read it, under terms that never mention your agreement. Whether a particular paste breaks a rule is a question for your general counsel. Our guide for bankers covers the bank side of the same problem.
What can an executive assistant paste into AI?
We sorted the material assistants handle by where it can go, using the terms above. "A personal AI" means a ChatGPT, Claude, or Gemini account your employer hasn't approved.
- The executive's inbox, calendar, and contacts: the employer's approved tools only. In a personal AI, describe the kind of email ("a polite decline to a conference invitation") with no names.
- Board decks, minutes, and committee papers: the approved tools or the board portal your company uses. Nothing from them in a personal AI.
- Deal code names, scheduling forms, and target companies: nowhere outside the deal team's approved systems.
- A partner's documents under an NDA: whatever the NDA allows, which usually means people with a need to know.
- The executive's personal matters (medical appointments, family, home address): keep them out of every AI unless the executive has asked you to use one for them.
- Your own templates and drafts with placeholders: an itinerary format, a meeting prep checklist, a briefing outline with [EXECUTIVE] and [COMPANY] in place of names. Any tool.
- Your own job: a raise request, a review, or a job search. Keep it out of the work account your employer can export.
A work laptop has its own monitoring, as our work laptop guide explains, whichever AI you open on it.
How Morse keeps an assistant's own drafts private
Every risk above comes from a copy or a grant, with the executive's material sitting on an AI company's servers under consumer terms, or a connector holding the same reach into the mailbox that you have. The executive's material belongs in the tools your employer approves. Morse is for what's yours: the template with placeholders, the email tone you're stuck on, and the wording questions that carry no names or identifying details.
Your chat history is stored in your browser, encrypted, not on Morse's servers, our servers don't keep your messages or log what you write, and Morse doesn't train AI on your content, sell your data, or show ads. You pick a privacy mode for each chat:
- Private, for templates and drafts with no names in them. New chats start in Private, on open-weight models like DeepSeek, Kimi, and GLM, plus xAI's Grok. Morse has zero data retention agreements with its inference, search, and other providers, so they've agreed to keep nothing, and where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't. Here's what zero data retention means.
- E2EE, for a draft you'd rather no one else read, with the identifying details left out. Your browser checks the secure enclave's hardware attestation, then encrypts the message so only that enclave can read it, and Morse's servers relay text they can't read. The enclaves are run by third-party providers, not by Morse, and the guarantee rests on the enclave hardware and attestation working as designed. E2EE runs open-weight models with no web search or image generation, and it comes with Pro, at $19 a month.
- Anonymous, to ask GPT, Claude, or Gemini how Outlook delegation works without the provider knowing it's you. Our proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider sees the text and may keep it for a limited time to check for abuse, under its own policy, so leave out your executive's name and your company's.
Because we don't store chats, we usually have very little to give authorities (Morse privacy). Clearing your browser or losing the device deletes your chats, and Morse can't recover them, so copy a finished template into your own files.
Try a chat from the Morse home page without an account, or start free with 15 messages a day and no card.
Common questions
Can my executive or my coworkers see my Copilot chats about a mailbox I manage?
Microsoft says Copilot Chat history from a shared or delegated mailbox is saved to your own primary account, and other people using the same shared mailbox won't see each other's Copilot conversations.
Does Gemini work in a Gmail inbox someone delegated to me?
No. Google lists Gemini in Gmail among the features delegates can't use, and its Workspace help says Gemini can't reach messages in an inbox someone delegated to you, only the ones in your own.
Will email I send for my executive show my name?
In Gmail, a delegate's own email address appears on mail they send from the account. In Outlook, mail sent with Send on Behalf shows both names next to From, and only an admin can grant Send As, which shows the executive's name alone.
Is a ChatGPT temporary chat safe for my executive's material?
It keeps the chat out of your history and out of training while it stays temporary, but OpenAI says it 'may retain a copy for up to 30 days for safety purposes,' under your personal account's terms. Material your employer controls still belongs in the tools it approves.