Private AI for startups: what happens to your pitch deck in ChatGPT, Claude, and Gemini

The Morse teamPublished November 3, 20267 min read

Private AI for startups depends more on your plan than on which app you use. As of October 2026, a founder using a personal ChatGPT, Claude, or Gemini account is under consumer terms that allow training on chats by default and human review of some of them, while the same companies' team plans promise no training by default and treat your content as your confidential information.

We read OpenAI's, Anthropic's, and Google's own terms, privacy pages, and pricing pages on October 24, 2026, plus the federal trade-secret statute, and put the consumer and business rules side by side below. This isn't legal advice. It's about who can see your company's information once you've typed it, and if a deal or a dispute turns on it, ask your lawyer.

Is it safe to put your pitch deck in ChatGPT?

On a personal plan, ChatGPT may train on it unless you opt out. OpenAI's pricing page marks Free, Go, Plus, and Pro alike with "Content is used to train our models" and "Opt-out available," and its page on consumer data asks: "Please do not enter sensitive information that you would not want reviewed or used." Our guide to whether ChatGPT trains on your data has the opt-out steps and what deleted and temporary chats keep.

ChatGPT Business runs the other way. OpenAI's enterprise privacy page (updated January 8, 2026) says "we do not train our models on your data by default" and that you own your inputs and outputs. On a small team, Business workspace admins "can view, access, export, and delete end user conversations," so a co-founder with admin rights can read your chats. OpenAI's own access still includes "specialized third-party contractors who are bound by confidentiality and security obligations, solely to review for abuse and misuse."

Does Claude treat confidential business information differently?

Only on Claude's business plans. Anthropic's consumer terms say its Commercial Terms do not cover "Claude.ai or Claude Pro use for individuals or entities," so a startup paying for Pro seats is still a consumer customer, with training on unless each person opts out. The Commercial Terms behind Claude Team and Enterprise say "Anthropic may not train models on Customer Content from Services" and "Customer Content is Customer's Confidential Information"; our Claude privacy guide covers the consumer settings and retention periods.

What about Gemini and Google Workspace?

Google draws the same line: on a personal account, the Gemini Apps Privacy Hub asks you not to enter "confidential information that you wouldn't want a reviewer to see," while Google's Workspace privacy hub (last updated October 6, 2026) says "your content is not human reviewed or otherwise used for Generative AI model training outside your domain without permission." Our guide to whether Gemini is private has the Keep Activity setting and Google's retention periods.

Consumer vs business terms, side by side

Every cell links to the company's own page, read on October 24, 2026.

The team plans have a floor. OpenAI's pricing page says "Business plans are available starting at 2 users," and Claude Team is "for teams of 2 to 150," at $20 a seat a month billed yearly (Anthropic). A solo founder either pays for an empty seat or works under consumer terms.

Do AI chats put your trade secrets at risk?

They can, because the law asks what you did to keep a secret secret. Under the Defend Trade Secrets Act, information counts as a trade secret only if "the owner thereof has taken reasonable measures to keep such information secret," and it "derives independent economic value" from not being generally known (18 U.S.C. § 1839(3)). The statute's list of covered information includes "financial, business, scientific, technical, economic, or engineering information," with "plans, compilations" and "programs, or codes" named outright. An unreleased roadmap, a pricing model, or source code fits that list. Pasting it into a service whose terms let it train on the text and let reviewers read it is something the other side in a dispute could point to. Whether it would sink a claim is a question for a lawyer, not for us.

Samsung restricted these tools after a leak in 2023. On May 2, 2023, Korea's KED Global reported that Samsung barred staff in its Device eXperience division from using ChatGPT, Google Bard, and Bing's chatbot on work computers, tablets, and phones, after an engineer "accidentally leaked internal source code, a business secret, by uploading it on ChatGPT" to check it for errors.

Can an NDA stop you from using AI on a partner's documents?

It can. Many NDAs let you share the other side's information only with people who need to know it and are bound by confidentiality obligations. OpenAI's own business agreement shows the pattern, allowing disclosure "only to its Affiliates, employees, contractors, and agents who have a need to know and who are bound by confidentiality obligations." An AI service on consumer terms has made no such promise to you. Before you paste a term sheet, a customer's data, or a partner's roadmap into a chat, check what the NDA you signed says about third parties.

How Morse keeps your company's numbers off AI company servers

Every row in that table describes a copy of your deck sitting on a company's servers, governed by whichever contract your plan happens to carry. Morse is a web app with over 30 models, open and closed, in one account, and it's built so your chat history lives on your device instead. You choose a privacy mode for each chat, by what the chat holds:

• E2EE, for the cap table, the model, and the roadmap. Type the figures that matter into a chat started in E2EE and pick an open-weight model like GLM, Kimi, or DeepSeek. Your browser checks the secure enclave's proof that it's genuine before anything leaves your device, then encrypts the message so only that enclave can read it. Morse's servers relay text they can't read, and the reply is decrypted only in your browser. The guarantee rests on the enclave hardware and attestation working as designed, and the enclaves are run by third-party providers, not by Morse. E2EE chats come with Pro, at $19 a month. Here's how end-to-end encrypted AI chat works.
• Private, for drafting the investor update. It's the default for a new chat. Open-weight models, plus Grok, run with zero data retention: Morse has zero data retention agreements with its inference, search, and other providers, so they keep nothing. Our explainer on zero data retention covers what that promise means.
• Anonymous, for asking GPT, Claude, or Gemini how a SAFE converts. Morse's proxy removes your account, email, and IP address and sends the message under Morse's own account, so the provider sees the question without knowing it's you. It may keep the message for a limited time to check for abuse, under its own policy, so keep your company's real numbers for E2EE.

In every mode, your chat history is stored in your browser, encrypted, and not on Morse's servers. Morse doesn't train AI on your content, sell your data, or show ads (Morse privacy policy), so there's no training switch to find. Because we don't store chats, we usually have very little to give authorities. Clearing your browser, losing the device, or deleting your account deletes your chats, and Morse can't recover them, so keep your own copy of anything you'll need for diligence.

Which setup fits which document?

• Generic questions (how a 409A works, what a pro rata right is): any assistant, or Morse's Anonymous mode to keep your name off them.
• Your own pitch deck draft: a team plan with training off by contract, or an E2EE chat on Morse.
• Cap table, financial model, source code, unreleased product plans: keep them out of consumer accounts; on Morse, type what you need into an E2EE chat.
• Someone else's documents under NDA: read the NDA first.

Try a chat from the Morse home page without an account, or start free and move to Pro when your numbers need E2EE.