The AI Setup Guides ยท Free preview

One full section, exactly as buyers get it.

This is Section 08 of the Claude Setup Guide, published in full with nothing trimmed and nothing gated. If the writing works for you here, the other sections read the same way.

From Edition 2026.3 of the Claude Setup Guide, version 3.1, last updated 19 July 2026. Prices, model names, settings paths, and links were checked against Anthropic's own pages on 16 July 2026. What each plan includes was not re-checked feature by feature, so where your screen and the guide disagree, the live pricing page wins.

Published in full

Privacy and data defaults.

Section 08

Privacy and data defaults 7 min

This is the most important page in the guide. Read it once and set it up, and you never have to worry about it again.

The default changed in 2026. On the consumer plans (Free, Pro, and Max), your new chats are used to help train Anthropic's models unless you turn that off. This is a reversal of how it used to work, so most people who signed up before are affected too.

How to opt out

  1. Open Settings, then Privacy

    You are looking for the training toggle.

  2. Find the "Help improve our AI models" setting

    That is what the toggle is called as of July 2026. The exact label has shifted before, so if the words differ, match the meaning rather than the wording.

  3. Turn it OFF

    That is it. Your chats are no longer used for training, and your data is kept for a much shorter window.

New signups, check the toggle during setup

You may be asked about this when you first sign up, with a big "Accept" button and a smaller toggle. Slow down for a second, find the toggle, and set it the way you want before you click through.

What each setting means for how long your data lives

ModeUsed for training?How long it is kept
Consumer, training ON (the default)YesUp to 5 years
Consumer, training OFF (you opt out)NoAbout 30 days
Incognito chatNeverNot saved to your history
Team, Enterprise, or APINo, by defaultShort retention (API logs about 7 days)

The private-mode escape hatch

The ghost icon at the top starts an incognito chat. It is never used for training and is not saved to your history, no matter how your global toggle is set. Use it for one-off sensitive questions you do not want stored.

One honest carve-out you must know

A policy update that took effect July 7, 2026 means conversations that get flagged by Anthropic's safety systems can still be used for training even if you opted out. Anthropic does not spell out exactly what triggers a flag. So opting out is good hygiene, but it is not a guarantee for truly confidential material. For that, use the business rule below.

The business rule of thumb

Consumer plan with training OFF is fine for personal and low-stakes work. For anything involving client data, staff data, or regulated information, use Team or Enterprise, where no-training is the default and the safety carve-out is not a concern. Never rely on the opt-out alone for confidential business data.

What is safe to paste or upload

Use this simple decision tree before putting anything into a consumer account.

Green, go ahead

Your own draft writing, public info, general questions, non-sensitive notes, a document you wrote and own.

Yellow, think first and redact

Business documents with names or numbers, internal notes. Strip anything you would not want seen, and turn training off first.

Red, do not, unless you are on a business plan with clearance

Passwords, full card or bank numbers, government IDs, client contracts, employee or HR records, health records, tax documents, anyone else's private data.

Do this now

Go to Settings, Privacy, and set the training toggle the way you want. If you handle client data, note that Team is your real answer and plan to move there.

The rest of it

Twenty numbered sections. In each guide.

You just read Section 08 of the Claude guide. There are nineteen more like it there, and twenty in the ChatGPT guide, plus the reference material at the back of both. Plan selection, account security, first wins, prompting, custom instructions, Projects, connected apps, the agent modes, workflow recipes, the prompt library, a 30-day plan, troubleshooting, a final checklist, and a plain-English glossary. Same voice, same level of detail, for Claude and for ChatGPT.

See what's included

$49 CAD plus tax, one time. Every future edition included, at the same link. 30-day refund, no explanation needed, and you keep the guides.