Guide 07 of 12

What Is the Best Private AI for Confidential Client Documents?

Short answer: There is no single best tool, because "private" can mean three different things: the document never leaves your computer, the provider keeps no copy after it answers, or the provider keeps a copy but promises not to train on it. Decide which of those your clients' documents need first. Local apps such as LM Studio give the first, a small group of hosted tools give the second, and the business plans from Microsoft, Google, OpenAI and Anthropic give the third.

Updated November 16, 2026

Start with where the document is allowed to go

Most confidential documents arrive with an expectation attached: an NDA, an engagement letter, a professional rule or simply a client's trust. Read that expectation as a limit on where the file may travel.

  1. The file must not leave our computers. Only a local AI app meets this. LM Studio says "Nothing you enter into LM Studio when chatting with LLMs leaves your device" (LM Studio). You trade convenience and model quality for it, and each laptop becomes part of your security setup.
  2. The file may go to a provider, but no copy may stay. You need a hosted tool that holds nothing after it answers, and you should read what it keeps for billing or logs.
  3. A provider may store the file, but under business terms. The major business plans fit here. They store your chats and uploads, under retention your admins set, and say they do not train on them by default.

Our privacy ladder sets out seven levels in more detail.

The options, by what they keep

Read on November 13, 2026, from each provider's own pages. Rows run from the most local setup to the most stored, not by quality. There are no ratings.

Tool Where documents go What it keeps Training
LM Studio (local) Your own computer, on the local-model path What it saves on that computer Not applicable on the local path
Faradex Each firm gets its own Workspace on dedicated infrastructure in the United States Files and conversations are held in memory and are gone when the Workspace is reset Currently Opus via AWS Bedrock, more information is available on our website. AWS does not store prompts or responses, and model providers have no access to them
Proton Lumo for Business Proton's servers; "Lumo is powered by open-source large language models" Chat history with "Zero-access encrypted chat history"; "Lumo never keeps your data or uses it to train LLMs, except with your express permission" As quoted
Microsoft 365 Copilot Your Microsoft 365 tenant "The stored data includes the user's prompt and Copilot's response" "aren't used to train foundation LLMs"
Gemini in Google Workspace Your Google Workspace Prompts and responses: "90 days to indefinite, as determined by admins" Not used for training "outside your domain without permission"
Claude Enterprise Anthropic's systems Chats and projects; "by default, projects are retained indefinitely" "No model training on your content by default" (Claude pricing)
ChatGPT Enterprise OpenAI's systems Conversations, for as long as "Your workspace admins" set "We don't train our models on your organization's data by default." (OpenAI)

Prices, for the plans that publish them: Lumo Professional is listed at "$14.99 $11.99 /user per month" (monthly and yearly billing) (Proton); Claude Enterprise is "US$20/seat/month, billed annually, 20 seat minimum", plus usage; ChatGPT Enterprise has no public price; Faradex is priced from $120 a month, no seat minimum.

Three mistakes to avoid

Treating "no training" as "no storage." Every business plan in the bottom half of the table makes the no-training promise and still stores your documents. See Zero data retention vs. no training.

Assuming the enterprise plan is the private one. ChatGPT Enterprise keeps conversations under admin-set retention. OpenAI describes zero data retention for its API; we found no such option for the ChatGPT app. Details are on Does ChatGPT Enterprise keep your data?

Forgetting your own copy. Whatever the AI tool keeps, your firm still holds the original document and probably the AI's output. Your own records policy covers both.

A quick way to choose

  • If the document may not leave the building, use a local app, and accept the hardware and setup work.
  • If it may go to a provider but should not stay there, shortlist the tools that hold nothing after use, and ask each one what it logs.
  • If storage under business terms is acceptable, use the plan your firm already pays for, set retention as short as your work allows, and keep personal accounts out of client work.

To compare team plans on price and seat floor as well, see Best private AI platforms for professional firms.

About Faradex. Faradex is a private AI workspace for small and mid-sized firms that handle confidential client work. Each firm gets its own Workspace on dedicated infrastructure in the United States. Files and conversations are held in memory and are gone when the Workspace is reset. AI model: Currently Opus via AWS Bedrock, more information is available on our website. AWS does not store prompts or responses, and model providers have no access to them. Pricing: from $120 a month, no seat minimum; see faradex.ai/pricing.