How to Stop Claude Training on Your Data
Since a policy change in autumn 2025, Anthropic may use your Claude conversations to train its models by default and keep them for up to five years. This guide shows step by step how to turn that off, and explains what the opt-out does and does not cover.
What this is about
On consumer Claude accounts (Free, Pro and Max), using your chats for model training is enabled by default. Unless you object, new and resumed conversations may flow into the training of future models and are retained for up to five years. With training turned off, the short 30-day retention applies again. If you enter sensitive or personal data, make that a conscious decision instead of leaving the default in place.
Step 1: Turn off model training in the settings
- Sign in to Claude and open Settings via your profile.
- Go to the Privacy section.
- Turn off the option “Help improve Claude”.
The setting applies to your whole account, on every device, and takes effect from the moment you change it: new conversations are no longer used for training. Important to know: an older chat that you continue counts as a new conversation. If you flip the switch late, avoid resuming sensitive old chats from the period when training was active. Data that has already flowed into a training run cannot be removed afterwards.
Step 2: Incognito chats for especially sensitive conversations
In addition, you can run individual conversations as an incognito chat. Incognito chats do not appear in your history, do not create memories and are never used for training, regardless of your other settings. Note: Anthropic may keep a copy for up to 30 days for safety reasons.
Step 3: Exercise your privacy rights via the Anthropic privacy center
Independent of the account setting, you can exercise your privacy rights directly with Anthropic: at privacy.anthropic.com you will find the privacy policy and the way to request access to your data or its deletion. Deleting individual chats in the interface also excludes them from future training runs.
Work, Enterprise and API accounts
For Claude for Work, Enterprise and Education, as well as the Anthropic API, inputs and outputs are not used for training by default. A separate opt-out is usually not needed there. If you use Claude professionally, such a plan gives you a more robust foundation than a consumer account.
Why this matters for EU users (GDPR)
Claude inputs quickly contain personal data: names, contact details, customer and contract information. Once such data flows into the training of an AI model, it can practically no longer be removed from it in a targeted way. That sits uneasily with core GDPR principles such as data minimisation (Art. 5 GDPR) and the right to erasure (Art. 17 GDPR). And if you enter third-party data in a professional context, such as customer or employee details, you remain responsible for it under data protection law. The training opt-out is therefore a sensible first step.
What the opt-out does not do
Important context: the opt-out only prevents training. Your inputs are still transmitted to Anthropic and processed and stored there in line with Anthropic’s privacy policy. Personal data you enter still leaves your area of responsibility.
This is exactly where AEGIS comes in: the browser extension detects personal data and pseudonymizes it before it reaches Claude, ChatGPT or Gemini. Training opt-out and pseudonymization complement each other. See how it works in the interactive AEGIS demo.
As of July 2026. Labels in the Claude interface may change; the current information in the Anthropic Help Center prevails.