Does Claude Train on Your Conversations? Anthropic's 2025 Policy Change Explained

For consumer Claude accounts (Free, Pro, and Max), Anthropic may use your conversations to train its models unless you opt out. This is a change from how Claude used to work. In August 2025 Anthropic updated its consumer terms so that new or resumed chats and coding sessions can be used for model training by default. Commercial products such as Claude for Work and the API are not covered by that change. This post explains what changed, how the setting works, and how to keep your own private record.

What changed in August 2025

Before the update, consumer Claude conversations were not used to train models by default. Anthropic then announced updated consumer terms and a revised privacy policy that flipped the default for consumer plans. Existing users were prompted to make a choice and had until October 8, 2025 to accept the updated terms.

The key facts of the change:

If you signed up or accepted terms after that window, the choice was presented during onboarding.

How retention differs based on your choice

Your training choice also affects how long Anthropic keeps your data:

• If you allow training: contributed data can be retained for up to five years.
• If you opt out: the retention period stays at 30 days for consumer accounts.

The change is forward-looking. Conversations already used in a completed training run, or one in progress, are not pulled back out. As with most platforms, the sooner you set your preference, the less of your history is ever eligible.

Why the policy changed, and what it means in practice

The shift reflects a wider industry pattern: real conversations are valuable training data, and most consumer AI products now use them by default unless the user opts out. The practical takeaway is not that Claude became less safe to use, but that the default flipped. Where consumer Claude conversations were previously excluded from training, the burden is now on you to set the preference if you want them excluded.

For day-to-day use this changes two things:

• You have to make an active choice. Leaving the default in place means new and resumed chats can be used for training.
• Your retention window depends on that choice. Allowing training extends retention to up to five years; opting out keeps it at 30 days. So the setting controls both whether your data trains models and how long it is kept.

How to opt out

The control lives in your Claude account settings under data and privacy. Look for the option that allows your data to be used to help improve Claude and turn it off.

A few details to keep in mind:

1. The setting applies to new or resumed chats and coding sessions, not to old chats you never reopen.
2. You can change it at any time, and the change applies going forward.
3. Opting out reduces retention to 30 days, so it affects how long your data is stored as well as whether it trains models.

Training, history, and memory are different things

As with other platforms, do not confuse three separate concepts:

• Training: whether Anthropic learns from your chats. Controlled by the data setting above.
• History: whether your conversations are saved for you to revisit in the sidebar.
• Project context: features that carry context within a Project, which is about usefulness in your own workflow, not training.

Opting out of training does not delete your saved conversations. It changes what Anthropic does with them.

How to check and confirm your setting

Because the default now allows training on consumer plans, verifying your setting matters more than it used to.

1. Open your Claude account settings and find the privacy or data controls.
2. Locate the option about using your data to improve Claude and confirm it is in the state you want.
3. Remember the scope: the choice applies to new or resumed chats and coding sessions. An old chat you never reopen is not used regardless.
4. If you change it, the change is forward-looking. Anthropic excludes your data from future training when you opt out or delete your account, but does not pull back data already used.

A practical note for coding sessions: if you use Claude through tools that resume or continue earlier sessions, those resumed sessions count as activity, so they fall under the current setting at the time you resume them.

Consumer plans versus commercial products

As with most providers, the strongest way to avoid training on your content is to use the commercial route rather than relying on a consumer toggle.

For teams handling sensitive material, the commercial tiers and the API change the default itself, which is a firmer guarantee than a per-account setting that can be reset.

Keeping a private, searchable record

Claude's native history, like most platforms, only searches conversation titles, not the full text of your messages. So even when your chats are saved, finding a specific answer from weeks ago is hard. Combine that with a 30-day retention window if you opt out, and a long-term personal archive is something you have to maintain yourself.

A local-first tool solves both problems. LLMnesia is a free, local-first Chrome extension that searches your AI chat history across ChatGPT, Claude, Gemini and 10+ platforms. It indexes your Claude conversations on your own device as you browse them, so the searchable record stays with you regardless of Anthropic's retention setting, and you get full-text search instead of title-only search.

For the equivalent breakdown on the other major assistant, see does ChatGPT train on your conversations?. For the broader picture, read are AI conversations private? and local-first AI tools and privacy.

Practical recommendation

For consumer Claude users doing real work:

1. Decide your training preference deliberately rather than leaving the default.
2. Remember that opting out also shortens retention to 30 days, so keep your own archive if you need long-term access.
3. Use a local-first indexer so your record survives the retention window and is actually searchable by content.