Does Claude train on my data?
Only if you allow it. On consumer plans (Free, Pro, Max), Anthropic uses your chats to improve its models only when you turn on the 'model improvement' setting, when a conversation is flagged for safety, or when you opt in. If enabled, data can be kept up to 5 years. Incognito chats are never used.
Why — the first-principles explanation
Training a model means feeding it example conversations so it learns better responses. Whether your specific chats become part of that pool depends on the product and your settings, not on some fixed rule that applies to everyone.
For Anthropic's consumer products (Claude Free, Pro, and Max), the company will use your chats and coding sessions to improve models only under specific conditions: you choose to allow it, your conversation is flagged during a safety review, or you explicitly opt in. There is a Model Improvement toggle in Privacy Settings that controls this, and an Incognito chat mode whose conversations are never used for training even if that toggle is on.
Retention is a separate question from training. If you allow your data to be used to improve Claude, Anthropic can keep it for an extended period, up to 5 years, because model work happens over long cycles. Feedback you submit (like thumbs up or down) is also retained up to 5 years. If you do not allow training, your data is kept under Anthropic's standard, much shorter consumer retention. Commercial products such as the API and Claude for Work follow stricter defaults and are not used to train Anthropic's models by default, which is why businesses can send sensitive material.
An example that makes it click
Think of a suggestion box at a restaurant. By default, the kitchen just reads your order, cooks it, and moves on; your note doesn't go into next year's recipe book. Only if you check a box that says 'you may use my feedback to improve the menu' does your note get saved and studied for a long time, up to five years. And if you write your note in a special 'private' slip (incognito), it never goes into the recipe book at all, even if you checked that box elsewhere.
How to do it
- Open claude.ai and go to Settings, then Privacy.
- Find the 'model improvement' or 'help improve Claude' toggle.
- Leave it off (or switch it off) if you do not want your chats used for training.
- Use Incognito chat for sensitive conversations you never want used for training.
- For work data, use Claude for Work or the API, which are not used to train models by default.
Key facts
- On consumer plans, chats are used for training only if you allow it, if flagged for safety review, or if you explicitly opt in.
- A Model Improvement toggle in Privacy Settings controls whether your data can be used to improve Claude.
- If you enable training, data can be retained up to 5 years; feedback data is also kept up to 5 years.
- Incognito chats are never used to improve Claude, even with Model Improvement turned on.
- Commercial products (API, Claude for Work) are not used to train Anthropic's models by default.
Anthropic's assistant, known for long-document reasoning and careful, safe answers.
Affiliate link — we may earn a commission at no cost to you.▶ The 60-second explainer (script)
Does Claude train on your data? Only if you let it. On the consumer plans, Free, Pro, and Max, Anthropic uses your chats to improve its models in just a few cases: you turn on the model improvement setting, a conversation gets flagged for safety, or you specifically opt in. There's a toggle in your Privacy Settings that controls this, so you can leave it off. If you do enable it, your data can be kept for up to five years, since training happens over long cycles. Want a conversation that's never used for training at all? Use Incognito mode. And if you're handling work data, the API and Claude for Work aren't used to train models by default. Check your privacy settings and choose what you're comfortable with.
What authoritative sources say
People also ask
Is training on by default for free Claude users?
Your data is used only if you allow it through the Model Improvement setting, if a chat is flagged for safety review, or if you opt in. Check your Privacy Settings to confirm your choice.
Does Anthropic train on API or business data?
No, not by default. Commercial products like the API and Claude for Work are not used to train Anthropic's models unless you specifically agree.
How long is my data kept if I allow training?
Up to 5 years. If you do not allow training, Anthropic applies its standard, much shorter consumer retention.
What is Incognito chat?
A mode whose conversations are never used to improve Claude, even if your Model Improvement setting is turned on.